Abdelati Elasri

How to Improve API Performance Pagination Pagination is a common optimization technique when result sets are large. By streaming results back to the client in pages, service responsiveness can be improved. Asynchronous Logging Synchronous logging writes to disk on every API call, slowing down the system. With asynchronous logging, logs are first sent to a lock-free buffer and control is immediately returned. The buffer contents are then flushed periodically to disk, significantly reducing I/O overhead. Caching Frequently accessed data can be cached for fast retrieval. Clients can query the cache first instead of hitting the database directly every time. For cache misses, the database can be queried as a fallback. In-memory caches like Redis provide faster data access compared to databases. Payload Compression Request and response payloads can be compressed using algorithms such as gzip to reduce transmitted data volume. This speeds up upload and download times. Connection Pooling Opening and closing database connections has significant overhead. Using a pool of open connections avoids this. The connection pool manages lifecycle events internally.

٢

Writing a compiler made me better at designing APIs. I built GALANT, a compiler for a French-based programming language. About 410 lines of C. Lexer -> Parser -> Semantic analysis. Why French? We learn in French here, but all programming languages are English-only. This creates barriers, especially for older teachers who struggle with English. GALANT lets you write: variable compteur = 0; tantque (compteur < 10) { afficher(compteur); } Instead of English keywords. What building a compiler taught me: 1. Abstractions have cost Every feature compiles to state machines. Understanding this helps you optimize. 2. Parsing is everywhere APIs parse JSON. Databases parse SQL. Same principles: tokenize, build a tree, validate. 3. Error messages matter "Syntax error line 37" is useless. "Expected ';' after statement" is helpful. Same applies to APIs. 4. State machines run the world Regex engines, protocol handlers, routers, all state machines. Would I use GALANT in production? No. Did it make me better at system design? Absolutely. repo: https://lnkd.in/dvGRbCPD #Compilers #SystemsProgramming #SoftwareEngineering

٣

I’m pleased to share a project we have been working on: 𝐒𝐞𝐢𝐬𝐦𝐨𝐅𝐥𝐨𝐰: 𝐋𝐚𝐦𝐛𝐝𝐚 𝐀𝐫𝐜𝐡𝐢𝐭𝐞𝐜𝐭𝐮𝐫𝐞 𝐒𝐲𝐬𝐭𝐞𝐦 🌍 This project demonstrates a production-style architecture combining streaming, batch processing, orchestration, observability, and strong reliability guarantees. with Salah Algamasy we designed and implemented a fault-tolerant, highly available data engineering platform capable of processing live global seismic events and joining them with static infrastructure data to generate real-time geospatial insights. 🔄 𝐏𝐫𝐨𝐣𝐞𝐜𝐭 𝐎𝐯𝐞𝐫𝐯𝐢𝐞𝐰: • Ingests real-time earthquake data from external WebSocket APIs • Streams high-frequency seismic events through a distributed message broker • Processes data using unified batch & streaming engines • Stores data in an ACID-compliant delta lake with versioning • Performs real-time stream–static joins with geospatial analysis • Computes proximity using the Haversine formula • Generates automated alerts for seismic events near critical infrastructure • Provides full system observability and automated orchestration 🏗 𝐀𝐫𝐜𝐡𝐢𝐭𝐞𝐜𝐭𝐮𝐫𝐞 𝐋𝐚𝐲𝐞𝐫𝐬: 1️⃣ Ingestion Layer — Real-time producers publishing structured events 2️⃣ Message Broker — Decoupled, high-throughput event buffering 3️⃣ Processing Layer — Stream processing with schema enforcement, and validation 4️⃣ Storage Layer — ACID-compliant Delta-based data lake 5️⃣ Batch Layer — Static geospatial airport reference data processing 6️⃣ Analytics Layer — Real-time proximity computation and alert generation 7️⃣ Monitoring Layer — Automated workflow management and full observability 🛠 𝐓𝐞𝐜𝐡 𝐒𝐭𝐚𝐜𝐤: • Apache Spark (Structured Streaming) • Apache Kafka • Delta Lake • Apache Airflow • Prometheus • Grafana • Docker & Docker Compose • Github Actions 📂 𝐆𝐢𝐭𝐇𝐮𝐛 𝐑𝐞𝐩𝐨𝐬𝐢𝐭𝐨𝐫𝐲: https://lnkd.in/dTfhS6bC We would appreciate any feedback or suggestions for improvement. #dataengineer #kafka #spark #deltalake #airflow #prometheus #grafana #CI/CD

١٦٩

‏٢٨‏ ‏تعليق‏

A picture is worth a thousand words: 9 best practices for developing microservices. When we develop microservices, we need to follow the following best practices: 1. Use separate data storage for each microservice 2. Keep code at a similar level of maturity 3. Separate build for each microservice 4. Assign each microservice with a single responsibility 5. Deploy into containers 6. Design stateless services 7. Adopt domain-driven design 8. Design micro frontend 9. Orchestrating microservices #micro #data #web #tools #ai #techniques

٤

During a previous internship at the company « AGI » in Algeria, I worked on improving an institute’s website called « INPED » that had several shortcomings, such as: -The design does not follow current trends, and navigation lacks smoothness. -There is an overload of informations and links, which can leave the user feeling disoriented or bored. -Insufficient mobile compatibility (non-responsive website). -Poor user experience: navigation issues and low accessibility. -The navigation menu is not visible when reaching the bottom of the page. -The site is poorly indexed by search engines and cannot be accessed directly through Chrome and is marked as potentially dangerous. -Some important functionalities are absent; notably, the registration for training, seminars, and internships is not digitized , it is still done through traditional methods such as phone calls or in person sign-ups. The following are the improvements I suggested : -Optimize the website's visibility on search engines and enhance the navigation to make it more fluid and user-friendly for accessing information. -Make the website more attractive and engaging. -Enhance the overall user experience. -Ensure the website is fully optimized for mobile and tablet devices. -Review the site structure and update content where outdated or incorrect information exists. **Add new features such as**: -Integration of an e-learning platform that allows users (learners) and trainers to register, publish, and access training programs. -User management: a registration and authentication system for administrators, users, and trainers. -Training catalog management -Online registration for training/seminars/internships (digitizing the process) Online payment system integration. -Enhanced data security and privacy It was a pleasure for me to contribute to this project (especially the frontend part) and to gain valuable experience in enhancing website functionality and user experience 🥰 #WebDevelopment #InternshipExperience #WebsiteRedesign #UserExperience #WebProject #HTML5 #CSS3 #Bootstrap #Laravel11

١

If you're into AI, you know how much time you can lose just trying to find the right learning materials. 😩 A big shout-out to #ChoroukMalmoum for sharing a fantastic set of resources that will save me (and hopefully many others) countless hours. What I love most is that 55% of the courses mentioned are on Coursera, where you can learn for free by using the audit option. This is perfect for students and anyone who wants to start their AI journey without financial pressure. Seriously, if you're an AI enthusiast, don't waste another minute searching—start learning from these right away! 🔥📚 1️⃣ The "AI For Everyone" course from DeepLearning.AI Level🎓 Beginner Best For : AI newcomers desiring a broad, non-technical overview of the field Link : https://lnkd.in/ebEu5wvG 2️⃣ Artificial Intelligence Nanodegree Level🎓 Beginner–Intermediate Best For : Anyone interested in learning a wide range of AI techniques from some of the top AI experts Link : https://lnkd.in/eSevyk5Z 3️⃣ Computer Science for Artificial Intelligence Professional Certificate Level🎓 Intermediate Best For : AI learners who want a solid computer science foundation Link : https://lnkd.in/eUZdgZMv 4️⃣ LangChain - Develop LLM powered applications with LangChain Level🎓 Intermediate Best For : Comfortable Python programmers who want to build LLM projects with recent advancements Link : https://lnkd.in/ePQ-yV8m 5️⃣ Large Language Models Professional Certificate Level🎓 Intermediate–Advanced Best For : Developers with machine learning experience who want to build more advanced LLM applications Link : https://lnkd.in/eRdfUQZf 6️⃣ Deep Learning Specialization Level🎓 Intermediate Best For : Students with some experience who want to dive into the deep learning branch of AI Link : https://lnkd.in/e52MKbAQ 7️⃣ Self-Driving Cars with Duckietown Level🎓 Advanced Best For : Advanced learners who have a strong interest in autonomous driving and enjoy hands-on learning Link : https://lnkd.in/eHh8UeJT 8️⃣ Artificial Intelligence Level🎓 Intermediate Best For : Self-starters looking for a completely free, top-tier course Link : https://lnkd.in/eTKHscCc 9️⃣ CS224N: Natural Language Processing with Deep Learning Level🎓 Advanced Best For : Intermediate to advanced students with ML experience who want to lay the groundwork for language-based systems, like LLMs, speach recognition, and information retrieval. Link : https://lnkd.in/e3A9HVN7 #Students #CollegeStudents #University #Education #OnlineLearning #FreeCourses #SkillUp #LearnAI #CareerDevelopment #AIforStudents #NLP #LLMs #GenerativeAI #DataScience #Coding #Python #LinkedInLearning #Coursera #Community

٣

I used to think learning Data Structures & Algorithms was just about solving lots of problems. But after a while I realized there are actually different levels to it. Something like this: Level 1 — You can code the algorithm if someone explains it. Level 2 — You understand why the algorithm works and its complexity. Level 3 — You can solve new problems by recognizing patterns. Level 4 — You can design or modify algorithms yourself. Most developers probably live around Level 2–3, and that’s already pretty strong. I made a small visual explaining these levels.

٧

Here’s my first Medium article — a guide on how to create a custom systemd service to run a Java Spring Boot application on Linux/Debian at boot time.

٢

‏١‏ ‏تعليق واحد‏

🚀 Final Year Project Completed! After months of hard work, collaboration, and late nights, I'm excited to share our Final Year Project — a complete end-to-end platform that allows users to train AI models automatically, without writing a single line of code. 🧠 What Does It Do? Our platform simplifies the machine learning workflow: Upload your own dataset Select the model type (classification or object detection yet) ✅ Classification Support: Choose from 45+ pre-configured models ✅ Object Detection: Train using the YOLO family of models Train the model with just one click Once training is complete, the system automatically provides: The trained model Confusion matrix Performance metrics Visualizations like accuracy/loss graphs This makes AI model training accessible even for users without deep technical backgrounds. 💻 Tech Stack Backend for model training: FastAPI Frontend: React.js Backend services: Express.js Database: MongoDB 👨‍🏫 Supervised by Jamal Shah 🙌 Special Thanks To my senior Yahya Younas for his constant guidance, and to my brilliant teammates Muhammad Qasim and Muhammad Usama Sharif — couldn’t have done it without you! We got a little delayed, but it was important to build a complete product — not just a prototype. And today, we’re proud to finally present our fully working system to the world. 🌍 #AI #MachineLearning #DeepLearning #FinalYearProject #FastAPI #ReactJS #MongoDB #ComputerVision #NoCodeML #StudentProject #FYP #ExpressJS

٦٣

‏١١‏ ‏تعليق‏

🚀 I’m working on a project called AUTOQAPRO – an AI-powered VS Code extension designed to make software testing faster and smarter. Right now, it supports JavaScript and TypeScript projects (.js, .jsx, .ts, .tsx). The tool can: Generate unit test files for any single code file in your project Ask your preference for a testing framework (currently supports Jest and Verbose) Automatically place the test file in your existing directory 💡 The idea is to reduce the time companies spend writing and maintaining tests. In the future, I plan to expand AUTOQAPRO to: Support integration testing Work with more languages Generate tests for an entire project (not just single files) The backend is built with FastAPI, deployed on Railway, and the extension itself is written in TypeScript using the VS Code extension framework. I’m also open for collaboration with anyone interested in AI, testing, or developer tools. Let’s build something powerful together!

٢٢

‏٢‏ ‏تعليق‏

Are your ML models in Tunisia suffering from data drift and inconsistent feature definitions across teams? You're not alone. The lack of centralized feature stores is a silent bottleneck slowing down many promising AI initiatives here. Think about it: multiple teams independently engineering the *same* features, leading to discrepancies and wasted resources. Imagine the efficiency gains if we could define, store, and serve features consistently across all models. We've seen significant improvements in model accuracy and deployment speed by adopting a feature store. It also simplifies regulatory compliance by providing a clear lineage of your features. It's about making data science more reproducible and reliable. What are your biggest challenges in managing features for ML in your organization? Let's discuss practical solutions! 👇 #MLOps #AIEngineering #DataEngineering #CloudComputing #TunisiaTech #FeatureStore #MachineLearning #Innovation MLOps AIEngineering DataEngineering CloudComputing TunisiaTech FeatureStore MachineLearning Innovation

١

For a long time, I used to say: "AI will not replace developers." Today, I see things differently. 🤖 AI isn’t threatening programming… It’s threatening the average developer. In Morocco and globally, the job market has changed. A company that used to need 10 developers can now operate with 2 or 3 — the rest is amplified by AI. The question is no longer: Will developers be replaced? The real question is: Who will remain? Those who stay will be the ones who: 🧠 Understand architecture, not just code ⚡ Use AI to multiply productivity, not resist it 📈 Understand business, not only technology 🔄 Learn faster than the market evolves The hard truth? The value of simply “writing code” is decreasing. 🎯 My personal view: This isn’t the end of the field — it’s a real filtering phase. Opportunities won’t disappear… but they won’t be for everyone. Evolve — or quietly become unnecessary. What do you think? Are we witnessing a revolution, or just a market correction?

١١

‏٢‏ ‏تعليق‏

Excited to share my latest project — Android Template Generator — now available as a Docker image🐳! This tool helps developers instantly generate ready-made Android project templates, saving setup time and improving productivity. Inspired by Spring Boot Initializr and Eran Boudjnah ’s Clean Architecture Plugin, it’s built to make starting new projects faster and cleaner. I’ve currently completed the basic framework, and I’ll be adding intermediate and advanced frameworks soon to further enhance the developer experience. 🔗 Check it out on Docker Hub: 👉 https://lnkd.in/gg8HXjh9 #AndroidDevelopment #DeveloperTools #Productivity #CleanArchitecture #Docker #AndroidTemplateGenerator

١٢