- Guest: Ran Reichman Co-Founder & CEO of Flarion
- Host: Gabor Szabo
- Language: English
- Location: Zoom
In this conversation we'll discuss the use of Rust at Flarion.
Ran is co-founder and CEO of Flarion, a company building high-performance data processing systems using Rust.
Prior to Flarion, Ran built high-scale data processing systems in cybersecurity, autonomous vehicles, and cloud infrastructure.
Transcript
1 00:00:02.840 --> 00:00:25.500 Gabor Szabo: Hello and welcome to the Code-maven live meeting. And if you're listening to the podcast then it's supposed to be the rustation station. If everything goes well, then, this conversation is going to be uploaded there as well. My name is Gabor Sabo. I usually help companies with rust
2 00:00:25.800 --> 00:00:45.769 Gabor Szabo: and python, and in this series of of meetings and conversations I try to discuss with guests how the the title is rost at work, so I'm trying to understand how people use rust at work, and
3 00:00:45.980 --> 00:00:51.359 Gabor Szabo: what are the considerations? What are the advantages? Maybe disadvantages? And so on.
4 00:00:51.610 --> 00:00:59.340 Gabor Szabo: And this time the 1st guest is Ron Reichman from Florian, Hi Ron.
5 00:00:59.720 --> 00:01:01.610 Ran Reichman: Hello, yeah. It's great to be here.
6 00:01:02.710 --> 00:01:06.595 Gabor Szabo: Thank you for accepting the invitation, and
7 00:01:07.530 --> 00:01:17.960 Gabor Szabo: let's get started. I think the 1st thing should be is that you you introduce yourself. We actually never met. So it's gonna be a 1st for me as well.
8 00:01:18.450 --> 00:01:25.170 Ran Reichman: Okay, excellent. Yeah. So thanks again for for having me here and excited for this chat.
9 00:01:26.269 --> 00:01:46.930 Ran Reichman: I'm Ron Reichman, a co-founder, and CEO with clarion. And what we do is at Clarion is we're a data processing company building solutions to accelerate and make data processing much more efficient, especially at large scale of the big data which means usually like
10 00:01:47.928 --> 00:01:53.750 Ran Reichman: terabytes, tens of terabytes, hundreds of terabytes. But we support a wider range of workloads.
11 00:01:53.930 --> 00:02:01.579 Ran Reichman: And yeah, we're definitely a rust shop and lots to talk about in terms of why rust, how we use rust, etc.
12 00:02:03.090 --> 00:02:14.739 Gabor Szabo: Can. Can you tell a little bit about your about your background? And this is a startup company, right? How long, how long? How long is the company? And what did you before? And how did you get there? That's that's sort of the.
13 00:02:16.080 --> 00:02:31.519 Ran Reichman: Yeah, it's a great question. So I'm born and raised in Israel and in the army. I was in the program which has recently become more famous, thanks to the whiz founders that are also graduates spent about 8 years
14 00:02:31.730 --> 00:02:37.570 Ran Reichman: in various intelligence units, low level development stuff like that.
15 00:02:38.384 --> 00:02:51.085 Ran Reichman: Then went on to work on a on a news startup briefly, and then spent about 4 and a half years in the in the United States and working on
16 00:02:51.900 --> 00:02:56.640 Ran Reichman: on autonomous vehicles and on cloud infrastructure.
17 00:02:57.424 --> 00:03:04.729 Ran Reichman: So I had a chance to work at Roblox, which has a data at a very significant scale, and before that, at embark trucks.
18 00:03:04.950 --> 00:03:09.610 Ran Reichman: and for the past 2 and a half, 2 years or so I've been working on what is now clarion.
19 00:03:11.050 --> 00:03:11.715 Gabor Szabo: Okay,
20 00:03:12.540 --> 00:03:18.980 Gabor Szabo: So it's flaring is is 2 years. And and when did you start using rust? Or when did you get
21 00:03:19.140 --> 00:03:26.519 Gabor Szabo: to start? Yeah with rust. Was it in Florian, or or already you started to use rust earlier.
22 00:03:27.520 --> 00:03:39.599 Ran Reichman: So I'm my. My background is C, plus, plus. I have a ton of ton of experience with the C plus plus. And when I started building Clarion. I knew that I was that I was interested in in using rust.
23 00:03:40.009 --> 00:03:46.669 Ran Reichman: Because primarily, I mean, there's there are several reasons that we can go into, but that that's when I started going.
24 00:03:48.270 --> 00:04:06.279 Gabor Szabo: Yeah, I'm definitely interested in in in. Why did you start? Why did you think that is good for for this company? And this is a startup company. So one of the things that I hear sometimes that it's not maybe not such a good language for startups.
25 00:04:09.420 --> 00:04:32.499 Ran Reichman: Yeah. So I I would disagree. I think I think it's a good language for startups. I'd say 3 main reasons that we're using spark that we're using rust here at the flare on the 1st is we're connecting to existing data infrastructure for our customers, and our customers have a very high expectations in terms of reliability.
26 00:04:33.603 --> 00:04:36.929 Ran Reichman: So it's a
27 00:04:37.596 --> 00:04:46.779 Ran Reichman: it's a it's a quite a significant advantage. The fact that we don't really have to worry about the memory, memory, corruption, and everything that that entails.
28 00:04:47.090 --> 00:04:49.459 Ran Reichman: so that that that was one thing
29 00:04:50.017 --> 00:05:06.560 Ran Reichman: the second is, there's quite a lot of interesting infrastructure today in the rust world. There's also a lot of great stuff in C, plus plus. But things like polars, like data fusion. These are open source projects that we were happy to plug into and integrate with.
30 00:05:06.760 --> 00:05:31.820 Ran Reichman: And the 3, rd which is an interesting one is that we kind of like the people who are into rust. So what we do at flare on is kind of it requires a lot of out of the box, thinking it requires trying out new systems, new ideas, and oftentimes the kind of people the kinds of people that are interested in the working in rust are also the kinds of people that we want to have at Larry on. Those were the 3 main reasons.
31 00:05:33.550 --> 00:05:34.460 Gabor Szabo: Okay.
32 00:05:34.520 --> 00:05:55.609 Gabor Szabo: Before I go, continue with this. Actually, I would like to ask something about this. I would like to mention that, unlike other podcasts or conversations, or whatever we have here, where there's only 2 people speaking here, we have a few other guests who are watching us and listening to us.
33 00:05:55.610 --> 00:06:13.769 Gabor Szabo: and I would like, and though they can't speak, but they can definitely ask questions in the chat. So I would like to invite everyone who is present that if you have any questions, just type in the question, and I'm going to take care, and we are going to. I'm going to ask those questions here as well.
34 00:06:13.790 --> 00:06:20.239 Gabor Szabo: So here already we have the the 1st questions from the from the audience is, Have you considered other languages.
35 00:06:21.470 --> 00:06:26.619 Ran Reichman: Yeah. So my, my, I definitely considered other languages. The default for me was to use the C plus plus
36 00:06:28.390 --> 00:06:42.050 Ran Reichman: I have a lot of good things to say about C, plus plus. I'm definitely not a hater of the language. I think so so much over the existing infrastructure. Of the of the software world was built on on top of C, plus plus
37 00:06:42.532 --> 00:06:49.617 Ran Reichman: but yeah, I think the 3 the 3 reasons that I mentioned or pushed us towards Ross.
38 00:06:50.730 --> 00:06:52.339 Ran Reichman: but yeah, and.
39 00:06:52.340 --> 00:07:08.300 Gabor Szabo: Okay, I'm actually especially interested in the 3rd one that you mentioned of the the type of people. I think that was the the 3rd one. So can you describe what? How would you describe this this type, or how do they? How do they differ from? I don't know
40 00:07:08.480 --> 00:07:09.360 Gabor Szabo: others.
41 00:07:10.050 --> 00:07:20.130 Ran Reichman: Yeah, so any generalization is wrong. Right? So there there are. It's it's hard to categorize people for sure. But what we've seen
42 00:07:20.240 --> 00:07:43.149 Ran Reichman: is that when people are curious about rust they're usually just curious people. They hear about a new technology, and they want to understand it better. And they want to try it out, and they wanna explore it. And we're a company fundamentally what we do like if if we zoom out is we look at it as like receiving high level instructions
43 00:07:43.753 --> 00:07:47.960 Ran Reichman: from our customers like this is what I wanna achieve with my data.
44 00:07:48.280 --> 00:07:54.179 Ran Reichman: And then under the hood, we, we might use different engines, different technologies, different techniques.
45 00:07:54.741 --> 00:08:09.439 Ran Reichman: So we're we're we have a very broad stack in terms of things that we need to understand and support. So we really want people who are curious about new technology and want to adopt a different and interesting technologies. Yeah.
46 00:08:10.460 --> 00:08:18.469 Gabor Szabo: Okay. So there's another question here which is actually jumping a little bit ahead, asking about Flurry and and get on, Github, which
47 00:08:18.960 --> 00:08:26.850 Gabor Szabo: it was sort of planned to be that I asked later on whether any of any part of Florian is open source.
48 00:08:27.330 --> 00:08:30.509 Gabor Szabo: and and if if he has done, where is it?
49 00:08:30.760 --> 00:08:31.890 Gabor Szabo: Hey? Welcome.
50 00:08:31.890 --> 00:08:32.789 Ran Reichman: Yeah, so
51 00:08:33.679 --> 00:08:51.449 Ran Reichman: excellent question. We're huge fans of open source. And we also build on top of open source. We don't currently have a public repository. We do, do a lot of contributions particularly to projects like a data fusion and arrow
52 00:08:52.087 --> 00:08:59.859 Ran Reichman: so that that's mostly our involvement in open source that like, we were heavy users of existing technologies. And we're able to push them to the limit
53 00:09:00.060 --> 00:09:13.719 Ran Reichman: and find interesting bugs, edge cases, etc. And yeah, and unpack them, unpack them, debug them and issue pull requests. So we have. We have quite a bit of those in the in the open source. Ecosystem.
54 00:09:13.720 --> 00:09:42.949 Gabor Szabo: If you're already started to go in this direction, then I have a few more questions, because I mean you mentioned 2 major projects that you're rely on major open source projects. But I guess you have tens, or maybe hundreds of of rust crates that use. And if you think about all the hierarchy that maybe thousands. And I'm really interested in. How do you deal with all the dependencies, the
55 00:09:43.550 --> 00:09:46.289 Gabor Szabo: all the problems that come with the dependencies.
56 00:09:47.620 --> 00:10:00.109 Ran Reichman: Yeah. So I don't think there's a good solution to this, like one of the challenges with rust, which isn't necessarily unique to us. But it's very you can feel it very strongly when you're building a a big rust project is
57 00:10:00.892 --> 00:10:21.569 Ran Reichman: the number of crates that grows quite rapidly you start with like 50. And then you're at 150. It's only 500. So yeah, definitely, lot of lots of dependencies. We invest in the Ci infrastructure so that the build is as fast as possible. We we're constantly maintaining
58 00:10:21.720 --> 00:10:24.008 Ran Reichman: the like the whole
59 00:10:25.040 --> 00:10:31.749 Ran Reichman: build set up in order for it to be as quick as possible. Sometimes. That also means using like powerful servers
60 00:10:32.212 --> 00:10:36.077 Ran Reichman: but I don't think there's a silver bullet here in terms of
61 00:10:36.812 --> 00:11:01.139 Gabor Szabo: What? Yeah. What I'm interested in is, how do you select crates? How do you make sure that they are? How how much can you invest in, or how much do you invest in verifying that they are secure enough that they performant enough? And how do you deal with with bugs, or any issues in these these crates.
62 00:11:02.440 --> 00:11:11.110 Ran Reichman: Yeah. So if if we encounter a a, we're quite picky so if if something is small enough, then we won't take a crate, but actually implement it.
63 00:11:11.574 --> 00:11:17.419 Ran Reichman: But in terms of bugs and reliability. So that's the that's the thing that I think we we invest more than anything else
64 00:11:17.740 --> 00:11:19.719 Ran Reichman: in terms of just like
65 00:11:20.614 --> 00:11:28.730 Ran Reichman: we really have tons and tons of tests to make sure that reliability is that is really as close as possible to 100%
66 00:11:30.390 --> 00:11:36.083 Ran Reichman: And that also is the case. When we introduce new crates, like, we're also we're we're we have
67 00:11:36.540 --> 00:11:56.890 Ran Reichman: a very expensive suite of testing for that in terms of a a security that's mostly a matter of reputation, like we don't choose crates that we think will do a bad job for us or that were that are shady. We look at like number of downloads. Are people using this in the ecosystem stuff like that?
68 00:12:00.420 --> 00:12:02.709 Ran Reichman: Yeah, I think that's mostly it.
69 00:12:03.020 --> 00:12:04.909 Gabor Szabo: Have you? Have you
70 00:12:05.070 --> 00:12:16.720 Gabor Szabo: considered or tried, or do you think that there is a need of financing, sponsoring certain open source developers who are
71 00:12:16.960 --> 00:12:26.660 Gabor Szabo: unrelated to your company, who maintain these these packages, or or bug bounties, or things like this.
72 00:12:27.540 --> 00:12:30.942 Ran Reichman: Yeah, I I completely, I think,
73 00:12:31.620 --> 00:12:35.570 Ran Reichman: I think companies like us should definitely invest in these kinds of things.
74 00:12:36.193 --> 00:12:39.720 Ran Reichman: We haven't yet detected like a
75 00:12:40.000 --> 00:13:02.590 Ran Reichman: the right projects in terms of the ones that we're actually heavily using and in need of these resources. Because we're pretty in the mainstream, like lots of Apache stuff, like lots of stuff that's already has a lot of investment. But yeah, if we encounter something that we can contribute to or even contribute financially, definitely, something that we'd consider. We we want the ecosystem to be sustainable.
76 00:13:03.920 --> 00:13:28.949 Gabor Szabo: Okay. So I think we can go back to to the mainstream sort of of this conversation. But actually, there are questions here that sort of related to this, which let me read it, because that's easier. Also, could you please elaborate a bit more on the nature of rust in your use case, what do you mainly do with rust? Do you use Asyncrust?
77 00:13:30.480 --> 00:13:33.583 Gabor Szabo: Yeah. So great question.
78 00:13:34.680 --> 00:13:42.290 Ran Reichman: The the nature of our use of rust is we use rust, basically in all of the data processing that we that we do
79 00:13:42.792 --> 00:13:50.970 Ran Reichman: so the way our system works is that we connect to existing infrastructure. So it can be something like written in Scala, for instance, Apache spark
80 00:13:51.885 --> 00:13:58.399 Ran Reichman: or in python and some of the ray infrastructure. And once we pass on
81 00:13:59.141 --> 00:14:05.909 Ran Reichman: the responsibility of the data processing from the high level language to a rust.
82 00:14:06.340 --> 00:14:18.049 Ran Reichman: We do everything in rust. We use a Jni when it's a Java virtual machine, and and python, we also use the relevant bindings to pass from
83 00:14:18.500 --> 00:14:28.769 Ran Reichman: from python to rust. And then it's basically rust all the way until we we have to give back the some sort of result. And then and then send it back.
84 00:14:29.470 --> 00:14:36.040 Gabor Szabo: So besides rust, what are what other languages do you use? Separate.
85 00:14:36.040 --> 00:14:37.139 Ran Reichman: Yes, we we use.
86 00:14:37.140 --> 00:14:42.359 Gabor Szabo: Python and and but I don't know if you're using actually or just somehow connecting.
87 00:14:42.360 --> 00:14:47.319 Ran Reichman: We're mostly connecting like we're using Scala and Python Scala more than python
88 00:14:47.896 --> 00:14:51.089 Ran Reichman: because of its use in Jvm environments.
89 00:14:52.790 --> 00:14:53.930 Ran Reichman: And
90 00:14:54.420 --> 00:15:06.250 Ran Reichman: but whenever we can implement something that's that requires performance, we do. We do it in rost. And that's most of what we do requires a a pretty significant performance investment.
91 00:15:08.060 --> 00:15:20.129 Gabor Szabo: Okay, so we can now circle back to the to the beginning of the company, and how rust was introduced to the company, and how and and how you
92 00:15:20.600 --> 00:15:35.980 Gabor Szabo: find new employees. I mean, this was a startup company. So there were. There was no one there. So it's not like a company where you have to start teaching people rust or so. It's totally your decision. If you are bringing in people with rust experience or
93 00:15:36.120 --> 00:15:40.115 Gabor Szabo: without, and then teaching them, how was that process?
94 00:15:40.850 --> 00:15:46.509 Gabor Szabo: What kind of people arrived? They were? Were they already experienced in rust.
95 00:15:47.640 --> 00:15:51.330 Ran Reichman: Yeah. So most of the people we hired are experienced in rust
96 00:15:52.128 --> 00:16:12.230 Ran Reichman: they have. sometimes that means just a general like generally, like as a hobby, have built things in rust. There's some crypto aspect because of the success of the Solana as a blockchain, so that that spurred a lot of rust development
97 00:16:12.762 --> 00:16:20.679 Ran Reichman: so we have a bit from there. And there are a few people who don't really have a rust background, but are a strong C plus plus developers.
98 00:16:20.870 --> 00:16:23.070 Ran Reichman: And there we were able to.
99 00:16:26.037 --> 00:16:46.762 Ran Reichman: there, there! It's more of a an investment as a startup. We don't do a lot of training. But if someone in our interview process shows potential and is able to learn rust relatively quickly, then we. We also consider those people. And we have a few of those. I'll say that
100 00:16:48.140 --> 00:16:58.174 Ran Reichman: we try just as a as a philosophy not to push rust to the limit in terms of performance. We push it to the limit in terms of
101 00:16:58.640 --> 00:16:59.560 Ran Reichman: and
102 00:16:59.990 --> 00:17:12.010 Ran Reichman: low level implementations, and we push it to the limit. But we don't like we don't try to be too sophisticated or too clever with the language. That's kind of my experience with C plus plus as well that like, it's possible to be very.
103 00:17:12.190 --> 00:17:33.960 Ran Reichman: very smart and very clever, but sometimes that makes the code harder to maintain. So if if one of you took a look at our code base, you'd see that it's it's mostly straightforward. And when we're doing something that's elaborate, it's usually because that was the last resort. And that's kind of my philosophy, regardless of rust, but especially in rust. Because you you can do a lot of very sophisticated stuff.
104 00:17:35.790 --> 00:17:47.790 Gabor Szabo: Can you elaborate a bit more? What are these things that you feel that there are too sophisticated, or or that you would like to that you're avoiding in in this case.
105 00:17:48.830 --> 00:17:50.340 Ran Reichman: Yeah. So I think,
106 00:17:52.090 --> 00:18:02.329 Ran Reichman: I mean, not. Not all of these are necessarily sophisticated, but the they add complexity. So when it comes to handling channels. Async. Things things in that area.
107 00:18:03.990 --> 00:18:31.159 Ran Reichman: sometimes sometimes. You have to use it. But if you can, if you can be like single core, running very powerfully and achieving your goals. Then we prefer that when it's possible that that's kind of the things that we're looking for. So anything that's like, I mean, this is this is the way in computer science generally, right? Like, once, once you introduce concurrency of various forms. Then things. Things become much more complicated the way when we can avoid that, we avoid it.
108 00:18:32.520 --> 00:18:59.869 Gabor Szabo: Right? So there was a question here that I'm not sure that too relevant to you. But maybe you have some input on this. So how easy for you to recruit developers! Our company is mostly C plus plus and python. And when I introduced Rust Code into the Python group for performance. I saw a lot of resistance from the developers, so I know that it's probably not, really. You haven't probably encountered this, but.
109 00:19:00.310 --> 00:19:08.610 Ran Reichman: I actually have some relevant experience, because I've I've interviewed a lot of people with C plus plus background. And I think there's a there's a divergence there are people who
110 00:19:09.303 --> 00:19:24.626 Ran Reichman: who really love c plus plus really appreciate what it's able to bring to the table. And they don't wanna switch to another language. And when we encounter those in our recruiting process, then we don't try to convince them, because they love what we're doing, and we're not doing what they're doing. And that's great
111 00:19:25.030 --> 00:19:33.180 Ran Reichman: and that that really, I think when when you're in an existing company and you want to introduce rust. I think. Actually, the Linux kernel
112 00:19:33.630 --> 00:19:36.849 Ran Reichman: did a great job in this regard when they're saying like.
113 00:19:37.070 --> 00:19:40.390 Ran Reichman: you can't, you don't really wanna migrate to rust
114 00:19:41.645 --> 00:19:56.979 Ran Reichman: cause, you really want to just migrate to anything basically in the software world, you wanna gradually add functionality show value. See? See that you're able to provide better, developer experience and improved velocity.
115 00:19:57.130 --> 00:20:01.643 Ran Reichman: and then you can encourage people to move on. But I think
116 00:20:02.190 --> 00:20:18.619 Ran Reichman: so if you're starting a new project. This is also what we're seeing in the data processing world. So if a company is using Apache spark and you say, Oh, no, you should use ray, or you should use duck. dB, usually, it's a tough sell because they're using a certain technology. And they like that technology. And they're proficient in that technology.
117 00:20:18.840 --> 00:20:26.089 Ran Reichman: And it's much easier to say, Okay, we're like in this new new team that's building a new technology from scratch.
118 00:20:26.320 --> 00:20:34.400 Ran Reichman: We can try something new. And I found that throughout my experience, and that that's much easier than moving a system from one technology to another.
119 00:20:34.961 --> 00:20:45.909 Ran Reichman: Yeah, that's also the case, like, regardless of rust and C plus plus like Java, go like golang. All of that stuff like there's there's ton of tons of examples in that regard.
120 00:20:47.520 --> 00:20:56.690 Gabor Szabo: The the people who are so. So you recruited a number of people who who basically didn't know rust, but were interested in in rust as well. Right.
121 00:20:58.663 --> 00:21:05.120 Gabor Szabo: What was their learning, process, or learning resources that you might recommend to to everyone else?
122 00:21:12.360 --> 00:21:17.029 Ran Reichman: I think 1st of all, the the Standard Rust Book.
123 00:21:17.350 --> 00:21:31.880 Ran Reichman: It's a good book. There's a lot of good stuff there. I think I read it cover to cover, like online, obviously. But I think I think when you want to get acquainted with rust and what it has to offer, then that's a good thing to use.
124 00:21:32.722 --> 00:21:39.955 Ran Reichman: I think when you're in learning mode, and you want to understand things, then,
125 00:21:41.140 --> 00:21:52.829 Ran Reichman: AI is quite useful like saying like, Hey, I know how to write this in c plus plus like, how would you write this in rust? And then you say, like, explain to me this decoration. Explain why it's not compiling stuff like that.
126 00:21:53.446 --> 00:22:06.289 Ran Reichman: Like AI has its issues and its challenges. And it's hallucination. But when when you're asking like, basic questions of like, this is what I'm doing. This is what I'm trying to achieve. This is how I know how to do it in Python C, plus. Plus. What have you?
127 00:22:06.400 --> 00:22:19.699 Ran Reichman: Then I found that that to be very effective, and I I do very much believe that it's easier to learn a programming language by doing so. Like reading a book is great, but you only internalize things when you
128 00:22:19.840 --> 00:22:29.000 Ran Reichman: and when you try to build, and the compiler is mad at you, and it's like, no, you can't do that. You can't do that either. You can't do that either, and rust in that sense is very annoying. It's a it's a strict boss
129 00:22:29.641 --> 00:22:34.760 Ran Reichman: but the result is good. The result is one in which you can really rely on the
130 00:22:35.310 --> 00:22:39.499 Ran Reichman: and the accurate accuracy, and
131 00:22:40.990 --> 00:22:45.530 Ran Reichman: the reliability of the result. I think that's kind of a Us.
132 00:22:46.830 --> 00:22:59.520 Ran Reichman: a switch that people need to do when they're moving to rust, and when they're trying out rust that, like the compilation, is hard, like getting things to compile is oftentimes the hard part. Then, when things run. They usually run accurately
133 00:23:00.252 --> 00:23:07.230 Ran Reichman: so that's that. That's what we tell people who are who are starting to learn. We're trying things out.
134 00:23:08.220 --> 00:23:08.970 Ran Reichman: Yep.
135 00:23:10.130 --> 00:23:29.490 Gabor Szabo: Yeah, so I do a lot of training. And my experience is that it's nice to explain things to people. But they actually really learn when they do the practice and when they get feedback for whatever they did, and well in rust the compiler gives a lot of feedback, so I have less work. But
136 00:23:29.660 --> 00:23:39.220 Gabor Szabo: still the human human feedback is nice when when you might already solve the problem. But there are better ways to to write in in the given language.
137 00:23:40.770 --> 00:23:45.740 Ran Reichman: Yeah, yeah, 100%. I also say, I think that once you're in a rush scheme
138 00:23:45.910 --> 00:23:56.859 Ran Reichman: which not everyone is in. And it's kind of a priv privilege. Then more junior members just ask a lot of questions, and our senior members like to like to share and mentor.
139 00:23:56.970 --> 00:23:58.060 Ran Reichman: So
140 00:23:58.511 --> 00:24:14.310 Ran Reichman: sometimes you're like, Well, why is this happening? And people just explain like, this is the philosophy behind how this works. So once you have some sort of a a core of a team that's able to provide these insights. Then continuously growing, it
141 00:24:14.780 --> 00:24:16.490 Ran Reichman: becomes easier.
142 00:24:16.490 --> 00:24:29.740 Gabor Szabo: Yeah. A nice part about this actually is that it's not not just not just the juniors who learn from such interaction, but also the more experienced people who encounter interesting
143 00:24:30.200 --> 00:24:45.140 Gabor Szabo: code snippets from the juniors. And then they have to have to think about why these people, why, that person wrote that kind of code. And and what's the story behind it? So I think it's it's in bi directional learning. There.
144 00:24:45.400 --> 00:24:51.879 Ran Reichman: Yeah, I think I think that's true about mentoring in general, like as a mentor, as a professor, as a teacher, you, you learn a lot from your students.
145 00:24:53.490 --> 00:24:55.330 Ran Reichman: I feel the same way as a manager.
146 00:24:56.880 --> 00:25:22.669 Gabor Szabo: Okay. So there is a question here which is actually very close to what I wanted to ask. But let's ask this 1 first.st So do you have any suggestion for someone who would like to find work in rust? Well, see, and other things, and is comfortable with them, but only has web, dev experience, career, wise career, wise. The transition seems borderline impossible.
147 00:25:24.830 --> 00:25:46.349 Ran Reichman: Yeah. So I I understand the sentiment of the comment. It's it's definitely not easy. Because the companies look at your background. And if your background doesn't seem related to what you're looking for I mean the the inconvenient truth is that it's just hard to break through. It's hard to say, like, Oh, yeah, this is this is what I previously did. But what now? I want to do something new.
148 00:25:46.830 --> 00:25:50.715 Ran Reichman: and I I think the
149 00:25:52.780 --> 00:25:59.919 Ran Reichman: they're the main solution to this, and it's not a convenient one, because it requires work. But that's why it's powerful
150 00:25:59.950 --> 00:26:27.659 Ran Reichman: is open source is contributions. Because, like, we have people that we interviewed for the company like we pursued them like we reached out to them that we never met. We don't know what else they did in the world. We just saw the kinds of contributions that they were able to make to interesting projects and say, this is the kind of person we want in the team, and we didn't know that they were like they could be Web Devs. They could be python they could be. I don't know what I honestly don't know what they're doing in their day job.
151 00:26:27.670 --> 00:26:47.530 Ran Reichman: but if you see people if they have like, it's kind of in that sense like art. So if someone shows you their art and the art is beautiful, then you say, like, Okay, I don't really care about the rest. I would like that someone on the team. And it's not easy, because you have to invest. And you have to like, learn and learn these environments, etc. You gain a lot from that. But you have to invest a lot of that.
152 00:26:47.550 --> 00:26:52.829 Ran Reichman: So yeah, not not not the the fast route or the easy route, but one that I think is effective.
153 00:26:53.780 --> 00:27:22.260 Gabor Szabo: Yeah. So so I'm afraid that you already answered my next question. But maybe you can elaborate a little bit more. So let I assume that as a startup you're growing so you're looking for more people. So if I wanted to join your company in the next months, too or so, so I don't have too much time. But let's let's assume I have some background that is relevant to you. What would you suggest to make
154 00:27:22.580 --> 00:27:24.339 Gabor Szabo: to make it
155 00:27:24.570 --> 00:27:35.679 Gabor Szabo: easier for me to to. Yeah, to make it easier for me to to find you or or you to find me, or or okay, whatever right? I think you understand the question.
156 00:27:36.700 --> 00:27:37.200 Gabor Szabo: I hope.
157 00:27:37.200 --> 00:27:39.179 Ran Reichman: Yeah. So 1st of all, the the.
158 00:27:39.180 --> 00:27:48.569 Gabor Szabo: To improve my chances. Actually, I think that should be the question. So what should I do to improve my chances that you actually find me, or employ me, and so on.
159 00:27:49.140 --> 00:28:00.230 Ran Reichman: Yeah, so it's a good. It's a great question. The easiest answer is, 1st of all to apply. that's that's the the 1st step, and we do look at applications. That that we get.
160 00:28:01.072 --> 00:28:04.150 Ran Reichman: I think. Given a certain application.
161 00:28:05.470 --> 00:28:09.219 Ran Reichman: You have to like our interview process.
162 00:28:09.682 --> 00:28:14.819 Ran Reichman: We. We work very hard for our interview process to be very similar to the work itself.
163 00:28:15.491 --> 00:28:36.339 Ran Reichman: I found that the at the like big tech, and most of the interviews are kind of IQ test. And it's not a bad thing like these companies like they they optimize for getting just very smart people, and obviously they're doing very well for with that. But for us, we want people who like during the interview really understand how their day to day is gonna look like.
164 00:28:36.440 --> 00:28:43.669 Ran Reichman: And they like it. They they see like they they do the interview task that we give them. And they say, Okay, this. These are the kinds of things that I want to do in the future.
165 00:28:43.840 --> 00:28:47.260 Ran Reichman: And sometimes people look at them and and they don't
166 00:28:47.370 --> 00:29:05.450 Ran Reichman: like they? They realize that they don't want to work on. I don't know low level optimization database. So what I would encourage like our website, Clarionio, you can see the kinds of technologies that we're using, which is like polars, data, fusion, arrow stuff like that.
167 00:29:05.580 --> 00:29:14.820 Ran Reichman: So if you engage with these projects and look at them and think how to use them. You don't even have to contribute to the open source. You can think like, Okay, how do? How do I use.
168 00:29:14.990 --> 00:29:16.870 Ran Reichman: and something like polars
169 00:29:18.430 --> 00:29:38.030 Ran Reichman: in data processing like, How how does that work? And if you're able to answer that question and share your experience and stuff like that. Then you're already ahead of, like, 90% of the people who who have no context don't know how any of these things work. And so oftentimes putting in the work puts you pretty much ahead.
170 00:29:39.170 --> 00:29:39.889 Gabor Szabo: Right?
171 00:29:40.980 --> 00:29:53.530 Gabor Szabo: okay. The next sort of the question of the expansion of this question would be, which is slightly, I feel strange with this question, because it sort of assumes a little bit of
172 00:29:53.891 --> 00:30:16.080 Gabor Szabo: moving ahead or or back in in time. So a little bit of time travel. But what would you suggest to someone if if the person has one or 2 years? So decides that, okay, I have no clue about rust and all of all of this stuff, but I know your company, and I really would like to work at at your company.
173 00:30:16.080 --> 00:30:26.009 Gabor Szabo: So I have. Now I'm going to start to do, learn things and do things. So in a year or 2 I have a much better chances to
174 00:30:26.710 --> 00:30:31.020 Gabor Szabo: get hired by Florian. Obviously you don't know what Florian will look like in 2 years.
175 00:30:31.020 --> 00:30:31.989 Gabor Szabo: Yeah, of course.
176 00:30:31.990 --> 00:30:41.889 Gabor Szabo: So that's why it's a little bit tricky question. On the other hand, you could think about it. What would you have suggested 2 years ago, or a year ago, for someone like this
177 00:30:42.040 --> 00:30:43.340 Gabor Szabo: to be hired now.
178 00:30:43.340 --> 00:30:50.260 Ran Reichman: Yeah, no, I I think that's a really interesting question to me. The answer is, if if you're looking at like the medium and long term.
179 00:30:50.520 --> 00:30:55.980 Ran Reichman: I think the my advice would just to be more ambitious. Like, if if you're saying like I
180 00:30:56.270 --> 00:31:08.050 Ran Reichman: for the next year or the next 2 years. I'm gonna work on this technology. Then you can really achieve a whole lot in that time, even as a side project. So I I would try to do something that excites me.
181 00:31:08.090 --> 00:31:28.381 Ran Reichman: Like 99% of the of the things that people work on. They just churn, they start working on it, and they they enjoy it for a while, and then they stop working for some reason. And that that's true for myself also, like that just happens like you start working on something, and then life happens, and you have a lot of other things to do. But if you're committed for a long time frame. Then you can really achieve a lot.
182 00:31:28.790 --> 00:31:36.020 Ran Reichman: and I would. I would just try to choose the something ambitious like, try to understand something, try to change something, try to do something like that.
183 00:31:41.010 --> 00:31:49.180 Gabor Szabo: Okay, yeah, I, I thought about some specific types of projects to work on or.
184 00:31:49.790 --> 00:32:01.669 Ran Reichman: So some people like, if we talk about specific projects like some people work on, for instance, the game engines and rust. If you're interested in gaming. There's this kind of joke that everybody wants to build a game engine in rust, but nobody wants to build a game.
185 00:32:01.900 --> 00:32:27.190 Ran Reichman: and so I don't think that's necessarily bad, but like these are the kinds of things like, if you're excited about gaming, try to build a game engine if you're excited about AI. Then try to see like what's relevant in AI with rust. I don't know if there's that much, but if you're excited about data, then you can look at these big data projects. There are quite a few at this point. Big rust projects online.
186 00:32:27.380 --> 00:32:50.029 Ran Reichman: And then I, I look at like the top 1520 rust projects on Github and say, like, what excites me here? What's interesting like, what would I want to build? What would I want to recreate. What would I want to benchmark or test, or stuff like that? And that also brings you connections like, if you, if you one of the things that I learned throughout my career is that when you reach out to successful people
187 00:32:50.614 --> 00:32:55.280 Ran Reichman: that are technical and you engage with what they're working on.
188 00:32:55.659 --> 00:33:03.900 Ran Reichman: Like, the response rate is huge. It's like, I don't know, maybe 90% or something like people people like that. People are excited about their technology and want to build on it.
189 00:33:04.420 --> 00:33:09.119 Ran Reichman: so finding those things that you actually want to engage with and actually engaging
190 00:33:09.685 --> 00:33:12.880 Ran Reichman: I I think is a is a great investment.
191 00:33:18.840 --> 00:33:21.160 Gabor Szabo: Yeah, that's that's an excellent point.
192 00:33:21.716 --> 00:33:24.769 Gabor Szabo: I I tried to Co, to
193 00:33:25.130 --> 00:33:29.710 Gabor Szabo: to recommend people all the time to to invest in open source
194 00:33:29.900 --> 00:33:43.889 Gabor Szabo: doing things in open source. But it's always a people get excited. But they they most of the people actually don't follow through this, and they do a little bit, and and then they abandon it the whole thing.
195 00:33:44.210 --> 00:33:50.790 Gabor Szabo: So those who do continue I guess they will stand out.
196 00:33:50.790 --> 00:34:02.239 Ran Reichman: Yeah, yeah, I think I think that's exactly right. I think because of the churn, and because most people don't go deep. So if you're able to go deep and and commit to something. Then again, you're you're ahead of, like 90% of people already.
197 00:34:07.240 --> 00:34:12.890 Gabor Szabo: So there's a comment here that for those a good start is trying to help in an existing open source project.
198 00:34:13.030 --> 00:34:14.480 Gabor Szabo: What do you think about that
199 00:34:16.440 --> 00:34:20.340 Gabor Szabo: open source, project or or create your new, create your own.
200 00:34:20.340 --> 00:34:36.189 Ran Reichman: I I wouldn't try to create my own as a start. But I think, even before helping even just using an open source project like. There's a huge one of the things that we're seeing at the flare on is that there's a huge variability
201 00:34:36.420 --> 00:34:39.879 Ran Reichman: in terms of what the open source project is able to do
202 00:34:40.510 --> 00:34:43.570 Ran Reichman: like some projects, especially in their core.
203 00:34:43.860 --> 00:34:55.739 Ran Reichman: They're super robust, they work, they, they, you just plug and play, and others are like they can be hard to use. They have a rough edges of various sorts. And being a user and understanding like.
204 00:34:56.170 --> 00:35:23.629 Ran Reichman: where does this technology really work. Well, where does it not work well, like? Why doesn't it work? Well? These are the kinds of things that you don't even have to start by like. Sometimes people are intimidated by contributing, because, like maybe people won't like what I contributed, or maybe I'll have a back and forth with some maintainer for days, and I won't get anything through. But just like starting with using and gaining insights and opening issues and saying, like, Hey, I saw this, is this. Okay, is this expected?
205 00:35:25.040 --> 00:35:36.049 Ran Reichman: again, these are the kinds of things that people just don't do. It sounds simple. It is simple, like it's not hard to take it to choose a project and and try to try to engage with it. But most people just don't do that.
206 00:35:37.400 --> 00:35:42.869 Gabor Szabo: Yeah. So a lot of a lot of this work is not very difficult, but time consuming.
207 00:35:43.170 --> 00:36:01.749 Ran Reichman: It's time consuming. It's an investment. There's no going around that. That's why it's valuable in a sense, like in the in the cryptocurrency world. There's a concept of a proof of work that you like show that you invested the time in something. I think, that's applicable to many, many aspects of life. So if you, if you invest that work, yeah, I think it has value.
208 00:36:03.060 --> 00:36:10.739 Gabor Szabo: Okay, so how do you evaluate the the decision that that you used? Rust?
209 00:36:11.040 --> 00:36:14.909 Gabor Szabo: What did you that you that you use your rust? Yeah. So.
210 00:36:15.610 --> 00:36:34.949 Ran Reichman: Yeah, it's a great question. It's a piercing question, I'd say, because it's it's hard to evaluate, right? You made such a big decision. And now you have to decide like, see if it's a good decision. I haven't regretted it yet, like I haven't seen a point where it's like, Oh, man, why did we do we use this technology? That was a huge mistake. We also have the very significant benefit of starting from scratch.
211 00:36:35.515 --> 00:36:42.215 Ran Reichman: So when you start from scratch. You don't have any legacy and before before the conversation we talked about
212 00:36:43.127 --> 00:37:00.820 Ran Reichman: like a migrating code to rust, so we don't really do that like we, we rewrite some things. And when you rewrite, then you're just like starting from scratch. So a lot of the pains of moving from one technology to another. We just skipped. It's much easier to start from scratch
213 00:37:01.857 --> 00:37:06.942 Ran Reichman: so I think I think it's too early to say for now like no regrets at this point.
214 00:37:08.110 --> 00:37:09.156 Ran Reichman: We'll see
215 00:37:10.100 --> 00:37:32.040 Gabor Szabo: Okay, what are the parts of of rust that or creates, or like the whole package? That you find exceptional good for for you or in general. And what? What are the areas where the rust world could improve? Where? It's something that hurts you.
216 00:37:33.280 --> 00:37:37.579 Ran Reichman: Yeah. So I'm I'll start with a positive I think
217 00:37:37.980 --> 00:37:54.099 Ran Reichman: rust in the past, say, 5 years has really matured in a in an impressive way. Like, you see, we engage with customers using all sorts of technologies, whether it's data, lake technologies or data processing or storage and stuff like that.
218 00:37:54.320 --> 00:37:55.320 Ran Reichman: And
219 00:37:56.448 --> 00:38:10.391 Ran Reichman: there's usually a crate for that, like the echoing the apples. There's an app for that. So usually there's a crate for that. Usually someone builds something pretty good that's able to use this technology, and and help
220 00:38:11.480 --> 00:38:13.060 Ran Reichman: help work with it.
221 00:38:15.330 --> 00:38:36.489 Ran Reichman: so that's something that I think is very good. I also think that the reliability is real like we. We did something that where we were integrating like Scala code and rust code. And suddenly there was like a memory corruption. And someone on our team was like, Oh, a memory corruption. I haven't seen one of those in years, so that that was like a moment of, Okay, yeah, this this rust thing is working.
222 00:38:36.650 --> 00:38:45.180 Ran Reichman: I'd say, the the 2 negatives that I can think of is one a build time
223 00:38:45.744 --> 00:39:05.519 Ran Reichman: there really is. It's a pain like building rust pretty quickly becomes pretty time consuming. And it's hard to optimize like when I when I worked in c plus plus, there were all sorts of tricks to to make the build run faster, even though I also encountered many situations where we had a very long build.
224 00:39:05.883 --> 00:39:23.279 Ran Reichman: But yeah, that that's that's a pain. And another one is, I'd say, the flip side of the there's a crate for that. But there's a lot of the support is that there's a lot of vaporware. I'd say there are like crates that you use them. And you're like, okay, this crate. It doesn't really work. And that's fine. I mean, you know, nobody like
225 00:39:23.620 --> 00:39:42.310 Ran Reichman: nobody's selling you that crate. It's just something that someone published online. But you realize that you have to. You have to prune it. You have to make decisions. You have to make sure that this this relates to your previous about, how do you choose the right crate? So you have to choose. You have to make sure that you're using something that you can believe in.
226 00:39:44.770 --> 00:40:12.500 Gabor Szabo: For a second, returning to the earlier conversation or a part of the conversation. We have another question here follow up to finding work as a web dev any suggestions for types of repos, types of projects that would be appreciated by employees employers if you contribute to. In your opinion, of course, I am contributing to programming languages. For example, I guess the
227 00:40:12.520 --> 00:40:17.530 Gabor Szabo: person means compilers. What do you mean? Programming languages.
228 00:40:23.480 --> 00:40:24.930 Ran Reichman: Yeah. So I think
229 00:40:25.750 --> 00:40:35.789 Ran Reichman: I think that's definitely good, like working on these kinds of technologies. It's not easy, something that like that that does differentiate you. But I will say
230 00:40:36.060 --> 00:40:43.710 Ran Reichman: that it's valuable to be able to like. I, currently, this is changing. But currently
231 00:40:44.217 --> 00:40:57.030 Ran Reichman: there's still a limited number of companies that are hiring for rust roles, and the the number is growing. But it's different. If you're coming to a company like aws, which does a lot of rust.
232 00:40:57.340 --> 00:41:25.589 Ran Reichman: or a company like Clarion that does a lot of us, just because, like, there are different use cases. So at Aws or Google, or something like that. You'll spend a lot of a lot more time on concurrency, because that's just like there's a lot of input, output happening inside their systems and a lot of micro services. So if I were interested in in a role in one of those companies, I'd probably try to understand like they have blog posts. They write interesting stuff about their challenges. I tried to understand a bit their stack
233 00:41:26.150 --> 00:41:27.080 Ran Reichman: and
234 00:41:27.591 --> 00:41:36.040 Ran Reichman: gain relevant experience. There, if we're talking about a company like Ferryon then, it's it's a different stack. So there are many islands in the rust world.
235 00:41:36.180 --> 00:41:38.988 Ran Reichman: and I I try to
236 00:41:39.890 --> 00:41:55.309 Ran Reichman: like, I think compilers are, are a great great place to start and if you find a a company building a compiler in rust, and obviously that's that's fantastic. But if you're looking at a company that's that's not doing exactly that. Then I would try getting acquainted with the with the tech stack that they're using.
237 00:41:56.210 --> 00:42:00.079 Gabor Szabo: Actually. For to this point I sort of
238 00:42:00.330 --> 00:42:17.380 Gabor Szabo: partially dreaming of about companies that would say to their to their candidates that, Okay, here are a couple of issues open in an open project, and your homework or home assignment for
239 00:42:17.380 --> 00:42:36.719 Gabor Szabo: for the interview process is to deal with those issues, and maybe not even directly related to the company. So to see how the person interacts with open source projects and and without and contributing to the world. Sort of
240 00:42:37.365 --> 00:42:37.660 Gabor Szabo: yeah.
241 00:42:37.660 --> 00:42:58.200 Ran Reichman: Yeah, I totally agree. I'd also add that the big advantage of engaging with projects that are related to the company is that you get a better understanding if you want to work there. This touches on what I mentioned earlier, like sometimes you, you might think that you want to work with. But then you look at like the the stack, what they're building, and you say, that that's not really my thing.
242 00:42:58.270 --> 00:43:09.230 Ran Reichman: And the the best time I. This is what I tell the candidates as well. The best time to understand that is, there is like, before you join the company. And cause. Yeah, it's really important to write that to find the right fit.
243 00:43:09.810 --> 00:43:28.770 Gabor Szabo: So here we have a very technical question or more technical than what we had. So far, high performance data data processing is is what you are. That's, I think the statement fly around inside. I don't know where this sentence comes from, so could you share any valuable practical examples
244 00:43:28.770 --> 00:43:44.370 Gabor Szabo: or tricks, how you debug distributed and high performance use cases. Do you use Tokyo for development, or just something like wrappers around arrow data, fusion, and so on. Any other valuable recommendations.
245 00:43:46.010 --> 00:43:47.834 Ran Reichman: Yeah, it's a big question.
246 00:43:49.980 --> 00:43:56.000 Ran Reichman: 1st of all, I mentioned this. But I'll mention it again, because it's just so important, like tests
247 00:43:56.250 --> 00:44:00.129 Ran Reichman: across the stack matter a whole lot. And
248 00:44:00.850 --> 00:44:07.189 Ran Reichman: this varies across companies. Because if you're building a like, I've I have some experience like with like a front end development.
249 00:44:07.370 --> 00:44:27.850 Ran Reichman: and there I don't know if I didn't invest a huge amount of my time in tests. You mostly want to iterate and see that it works and find bugs and solve them. When you're building a mission critical software, which is what we're doing. You really want things to be exact, like some startups can move fast and break things at play around. We don't move. We try to move as fast as possible, but
250 00:44:28.350 --> 00:44:51.179 Ran Reichman: it really needs to be stable. So this means integration tests, system tests, unit tests. These are the kinds of things that that are very useful to debug even distributed and high performance systems, because even a distributed system eventually includes various components and interactions between those components. And if you test those well.
251 00:44:51.380 --> 00:44:54.490 Ran Reichman: then you reduce the number of cases that said
252 00:44:54.620 --> 00:45:01.289 Ran Reichman: our team finds itself running something for 3 h on 2 TB of data in order to find all sorts of edge cases.
253 00:45:01.630 --> 00:45:02.950 Ran Reichman: And there.
254 00:45:03.110 --> 00:45:04.190 Ran Reichman: I think.
255 00:45:05.150 --> 00:45:12.210 Ran Reichman: 1st of all, it's not something that we've solved, I'd say, and we're working hard on building tooling.
256 00:45:12.650 --> 00:45:37.429 Ran Reichman: That extracts data from the data processing pipeline and visualizes it in a way that's easy to understand. So that you understand, like, how much memory you're using, how much CPU like, why is it happening? How much disk? And interestingly, this is something that is also gonna be valuable for our customers. So like, we make data processing much more efficient, much faster, much more performant.
257 00:45:38.376 --> 00:45:49.580 Ran Reichman: But by doing that we also wanna visualize this to our customers. So in a sense, we wanna visualize it to ourselves to understand what the bottlenecks are, and then visualize it to our customers. This is a hard thing to do.
258 00:45:51.640 --> 00:46:01.180 Ran Reichman: so I I would say, we we've solved it. But I'd say, adding a lot of debugging. We use also a a Tracy
259 00:46:01.360 --> 00:46:03.801 Ran Reichman: actually, as a tool to
260 00:46:04.570 --> 00:46:11.639 Ran Reichman: defined performance issues, and it's quite effective so when you're able to attach things to your code.
261 00:46:11.800 --> 00:46:14.280 Ran Reichman: and they do all sorts of
262 00:46:16.100 --> 00:46:24.969 Ran Reichman: performance, monitoring of various sorts and benchmarking we found that to be valuable, especially when when debugging performance issues.
263 00:46:27.390 --> 00:46:47.120 Gabor Szabo: So the next question that coming from the audience is actually the next question I wanted to ask. So I'm going to ask in the way I thought about sort of is that to try to understand what is your experience? Using AI tools to write rust code, which
264 00:46:47.770 --> 00:46:51.410 Gabor Szabo: tool you might be using, which model.
265 00:46:52.960 --> 00:46:59.420 Ran Reichman: Yeah. So 1st of all, let's say, we're model agnostic at Flareon I found that the AI
266 00:46:59.910 --> 00:47:19.169 Ran Reichman: it's similar to ides and even operating systems. People have their preferences. And as a as a manager, I I'm not gonna push someone to use a certain flavor of AI, because because I have a preference for it. Like, I trust the the team members to
267 00:47:19.650 --> 00:47:26.699 Ran Reichman: try to be as effective and efficient as possible. And we bring people that like are trustworthy in that regard.
268 00:47:27.287 --> 00:47:42.000 Ran Reichman: I've had good experience with Cloud. I think it's it's good for software development. Some team members really like a chat gpt some like deep seek they found, especially our one to be effective.
269 00:47:43.830 --> 00:47:45.410 Ran Reichman: I'll say that the things that
270 00:47:45.540 --> 00:47:50.200 Ran Reichman: the AI is best at in terms of our development is,
271 00:47:52.070 --> 00:48:06.599 Ran Reichman: 1st of all, needle in a haystack issues where you have like this huge error. And you say, like, Hey, AI! Like, what? What do you think is going on here? And then it's like, Oh, you! You wrote this incorrectly, and you should fix it. And usually, usually that's just right. Like you take a blob of code. You take the error
272 00:48:07.190 --> 00:48:11.500 Ran Reichman: 9 out of 10 times. It says like this is probably there. And and it's right.
273 00:48:13.520 --> 00:48:42.860 Ran Reichman: basic functionality. Is good. But usually that doesn't save that much money. I think the metric Sam Altman has said, I said this, and I 100% agree. I think the metric of like percent of code written by AI is is what in startup world they call a vanity metric which is like it's. It sounds interesting, but it doesn't really tell you how much how much you're using. Because, like autocomplete, like, how much of my text that I write in emails is written by autocomplete? Probably a lot. But like, how significant is that to my
274 00:48:43.170 --> 00:48:45.649 Ran Reichman: in my process, and not too much.
275 00:48:48.040 --> 00:48:53.059 Ran Reichman: I will say that where AI isn't as good is when things get complex.
276 00:48:53.310 --> 00:48:55.449 Ran Reichman: and that happens a lot for us.
277 00:48:55.880 --> 00:49:02.270 Ran Reichman: If you're trying to write something very performant. If you're trying to write something that's distributed and
278 00:49:02.390 --> 00:49:27.841 Ran Reichman: different things impact each other, whether it's memory. CPU disk, usage networking. Once many parameters get into the equation. Usually the AI focuses on some local minimum and not on like the whole system, and we found it to be very difficult to get it to solve. Those kinds of problems which is kind of fun. Because, like you want to be working on things that are
279 00:49:29.031 --> 00:49:34.358 Ran Reichman: not like the AI has trouble with, because otherwise, I mean, I don't know. You can just get the AI to do them. So
280 00:49:37.560 --> 00:50:06.820 Gabor Szabo: Let's switch a little bit to a slightly different audience. The audience, that part of the people who are here. I guess they're part of that group. If you're talking to other startup founders, or Ctos or so manager level people, or or co-founders, what would you recommend them? How to decide whether to use rust
281 00:50:06.870 --> 00:50:08.860 Gabor Szabo: or or not to use rust.
282 00:50:11.550 --> 00:50:12.220 Ran Reichman: Yeah, so.
283 00:50:12.220 --> 00:50:15.819 Gabor Szabo: How to, how should they evaluate it, or.
284 00:50:17.270 --> 00:50:25.230 Ran Reichman: So the place I'd start is, you want to have some familiarity with it. You like you want to champion. I think that's true with any new technology.
285 00:50:26.069 --> 00:50:29.501 Ran Reichman: You don't wanna like if you have.
286 00:50:31.026 --> 00:50:40.180 Ran Reichman: an engineer with the 3 years of experience, and they say, like, hey? Rust is exciting. I want to bring this to the company. You're unlikely to be successful by doing that.
287 00:50:40.560 --> 00:50:42.820 Ran Reichman: but if you have someone who can.
288 00:50:43.080 --> 00:50:50.090 Ran Reichman: who can experiment. Who can mentor who can make decisions on the right projects for this technology?
289 00:50:50.879 --> 00:51:01.740 Ran Reichman: Then I think you're much more likely to be successful. And I mentioned this earlier, but I'll mention it again, because I really think it's it's crucial, is choosing the right projects
290 00:51:02.040 --> 00:51:09.050 Ran Reichman: is super super important. So if you have like a contained project that you know what you want to achieve there, and you're starting from scratch.
291 00:51:09.230 --> 00:51:17.280 Ran Reichman: and you want to do that with rust, and you have the the know how, then I don't see why it shouldn't succeed like it's a very much set up for success.
292 00:51:17.400 --> 00:51:20.084 Ran Reichman: But the more the more the project is
293 00:51:20.720 --> 00:51:28.519 Ran Reichman: distanced from that situation, I think the less likely it is that that, it'll it'll be a success.
294 00:51:32.260 --> 00:51:33.280 Gabor Szabo: Okay.
295 00:51:34.410 --> 00:51:44.839 Gabor Szabo: I am, I think, mostly at the end of of my questions. I don't know if the audience have more questions. The audience who is here, or if you have the
296 00:51:45.320 --> 00:52:02.240 Gabor Szabo: subjects that you would like to bring up, that I I didn't ask because I didn't think about that important, and you think it would be important to for people to know about the company, about the use of rust, how to apply for a job at you to your company.
297 00:52:03.370 --> 00:52:06.200 Ran Reichman: Yeah. So how how to apply?
298 00:52:06.320 --> 00:52:19.700 Ran Reichman: It's pretty easy, like we have linkedin posts at our company, so you can look at Linkedin, at at flare on and see probably the most relevant here at the low level developer, which is how we call our rust position.
299 00:52:21.420 --> 00:52:27.170 Ran Reichman: Yeah, I think there's a lot of exciting momentum in the rust world. Lots of talent.
300 00:52:27.370 --> 00:52:30.350 Ran Reichman: I think the fact that big tech is investing in rust
301 00:52:30.570 --> 00:52:36.710 Ran Reichman: is really significant. Because, it just creates
302 00:52:37.390 --> 00:52:48.250 Ran Reichman: a whole set of people who are being trained and using the rust in production, and they also contributing to open source projects. So I think that's great.
303 00:52:48.870 --> 00:52:55.240 Ran Reichman: And yeah, lots of exciting stuff. Here, we're we're in terms of Larry on. So we're
304 00:52:56.010 --> 00:53:14.129 Ran Reichman: like, the world of data is so vast. And there's a there's so much to do here, and so many engines and technologies that can be applied when a company wants to do data processing very efficiently. So we're obviously excited about that future. And we think about rost has a has an important role. There.
305 00:53:16.500 --> 00:53:20.470 Gabor Szabo: Actually, I just thought about a new question, what are the main
306 00:53:20.890 --> 00:53:33.290 Gabor Szabo: problems that Florian is is facing that your engineers will need to solve in the foreseeable future that the new people who are joining the team might need to work on.
307 00:53:34.600 --> 00:53:43.313 Ran Reichman: Yeah. So I mean, I'd say, our engineers are pretty fortunate to have pretty tough challenges. so
308 00:53:44.160 --> 00:53:51.380 Ran Reichman: in in many senses. We're building a a distributed database for data processing, and that means
309 00:53:51.710 --> 00:53:55.649 Ran Reichman: handling issues of a very significant scale
310 00:53:55.900 --> 00:54:12.259 Ran Reichman: of a disk usage of memory management. these are the kinds of things that we're we're excited about like, how much, how much CPU is this using? How much memory is this? Designed correctly. Should we redesign this part of code of the code, should we?
311 00:54:14.072 --> 00:54:19.147 Ran Reichman: sometimes we, we insert many fast paths into our code, so that,
312 00:54:20.238 --> 00:54:25.790 Ran Reichman: when a customer is using is using it naively, and we can get a big performance boost
313 00:54:26.350 --> 00:54:39.550 Ran Reichman: so lots of challenges in that area, it's like CPU memory large scale of data and and performance. Those are the kinds of things that we encounter basically every day.
314 00:54:39.790 --> 00:54:49.780 Ran Reichman: And we also encounter a lot of like new and interesting technologies because our customers have a lot of new and interesting technologies that we we have to learn about them support them, etcetera.
315 00:54:50.850 --> 00:54:55.979 Gabor Szabo: Actually, who? Who are your customers, or what type of of companies are? Are your customers, or
316 00:54:56.750 --> 00:54:58.330 Gabor Szabo: could be your customers.
317 00:54:58.900 --> 00:55:04.656 Ran Reichman: Yeah. So our our customers are mostly the enterprise. meaning the companies that
318 00:55:05.478 --> 00:55:13.230 Ran Reichman: that are very significant data infrastructure. Because they're the ones that carry them care the most about performance. So if the company has a
319 00:55:13.360 --> 00:55:19.639 Ran Reichman: hundreds or thousands of data processing jobs, whether it's a spark array, or hadoop, or trino.
320 00:55:19.830 --> 00:55:27.419 Ran Reichman: These are the kinds of companies that really care about performance that really have pains in terms of managing their spark and their data infrastructure in general.
321 00:55:27.590 --> 00:55:32.639 Ran Reichman: And they're they're the ones that that are most excited about what Flarion brings to the table.
322 00:55:34.620 --> 00:55:41.769 Gabor Szabo: Okay, probably that should. That shouldn't have been the the last question. But so if somehow, this is how it turned out, anyway.
323 00:55:41.770 --> 00:55:42.170 Ran Reichman: Okay.
324 00:55:42.420 --> 00:55:50.540 Gabor Szabo: I don't see any more questions from the audience any any final words for this meeting podcast.
325 00:55:51.180 --> 00:55:58.989 Ran Reichman: No, I mean, I I appreciate this conversation and thanks everyone for joining and asking interesting questions. And yeah, thank you, Gabor, for organizing.
326 00:55:59.380 --> 00:56:22.929 Gabor Szabo: Thank you very much, and I really enjoyed it. And if you're watching this as a video, then please like the video. And if you listen to on, on the podcast, then in the show notes, there are going to be links where you can find the company and the other things that we discussed. And thank you, everyone for. Thank you, Ron, to
327 00:56:22.930 --> 00:56:30.949 Gabor Szabo: be here, and everyone else who joined us and asked questions that were enhanced. I think the the whole conversation.
328 00:56:31.000 --> 00:56:33.560 Gabor Szabo: and thank you everyone for listening.
329 00:56:33.960 --> 00:56:34.890 Gabor Szabo: Goodbye.
330 00:56:35.260 --> 00:56:35.960 Ran Reichman: Thank you.
