Rajkumar D

A lot of companies treat data like a smoke alarm: - It sits quietly in the background - Annoying when it chirps - Ignored when it works - Disconnected when it’s inconvenient Until something catches fire. Then, suddenly, everyone wants to know why no one heard the alarm. In my early years of working in data/analytics/AI - I noticed a strange organisational behaviour: We (and this included me) often only cared about data when it was too late to act on it: - Customer churn at a Telco: No one looked at the signals until revenue fell off a cliff and call centre volumes spiked - Data pipeline error at an Insurance Broker: No one fixed it until the board report was wrong (twice) - Quality issues in Yellow's mobile app: No one tracked them until the App Store rating dropped below 3 stars And then... data became the only thing that mattered. “What happened?” “Why didn’t we see this?” “Who owns the metric?” “Can we get a dashboard on this by tomorrow?” A lot of companies don’t have a data problem. They have a timing problem. - They wait for pain before they invest - They wait for loss before they measure - They wait for the post-mortem before they act If you treat data like a tool for accountability, it will always show up after the mistake. But if you treat data like a partner in decision-making, it starts working before things go wrong. And that is where the real value of data lives - not in the insight, but in the interruption. You don’t buy a smoke alarm because your house is on fire. You buy it because you want time to respond before it is. Your data systems should be the same. Not reactive. Not performative. But preventative. Because once things go up in flames, the only question that matters is: Why didn’t we see this coming?

63

14 Comments

So I built a small tool. It fetches real-time NAV data across AMCs and turns it into clean, analyzable datasets. Current version: • Tracks 600+ mutual funds • Calculates 1Y, 3Y, 5Y and since-inception returns • Exports structured data for analysis Stack: HTML • JavaScript • JSON APIs Data source: NAV feeds + MF APIs+Factsheet. Next version: • Fund rankings • Category filters • Risk metrics • A mutual fund recommendation engine based on investor surplus and allocation. This is Version 0. Building tools forces you to understand markets. Data → Insight → Better investing. #FinTech #MutualFunds #BuildInPublic #DataTools #Investing https://lnkd.in/dAmjzngq (click on fetch all funds or go for specific fund house)

79

5 Comments

The Data vs. Reality Gap in Mumbai’s Air Quality (More Data vs Better Data) Last night, I just walked 100 meters in Mumbai and witnessed a "scientific miracle." According to three different air quality monitors in that tiny radius, the PM2.5 levels were: 📍 Monitor A: 28 (Fresh mountain air?) 📍 Monitor B: 118 (Moderate haze) 📍 Monitor C: 188 (Unhealthy) (Surprisingly all 8s 😶‍🌫️) In the span of a two-minute walk, I apparently moved through three different climate zones. Mumbai has made a massive push to install these indicators across the city, and from a smart-city perspective, that’s great. But data is only as good as its reliability. Why does this matter? - Public Trust: If the numbers feel "random," people stop looking at them. - Actionable Insight: How do we decide when to wear a mask or keep kids indoors if the margin of error is over 400%? - Policy: We can’t fix what we can’t accurately measure. Hyper-local factors like construction dust or vehicular exhaust can cause spikes, but a gap this large suggests a need for better calibration and maintenance of our urban sensor networks. We have the hardware. Now, we need the accuracy. Has anyone else noticed these massive discrepancies in your neighborhood? Is "More Data" always "Better Data"? #Mumbai #AirQuality #SmartCities #Sustainability #DataIntegrity #UrbanPlanning #Pollution

33

ElevenLabs’ Scribe v2: Real-Time Transcription ElevenLabs has unveiled Scribe v2 Realtime, capable of live transcription in 90+ languages with 93.5% benchmark accuracy. With ultra-low latency and regulatory compliance, Scribe v2 empowers developers to build next-gen meeting, captioning, and voice applications. #ElevenLabs #VoiceAI

6

🔎 Hiring Panel Lens [Series 19] Why Strong PMs Sound Junior In AI Interviews After multiple AI-focused interviews, one pattern is very clear: Strong PMs are getting downgraded… not because of capability — but framing. When asked about AI work, most candidates start with: ➡️the model ➡️the feature ➡️the implementation Hiring panels are listening for something else: 👉 What decision the system is responsible for 👉 What happens when that decision is wrong 👉 How you designed for that failure Example: Weak answer: “We built an LLM to automate support.” Stronger answer: “We needed to decide which support queries could be resolved autonomously without increasing escalation risk.” Same system. Very different signal . AI products are not model problems. They are decision systems under risk. Quick self-check: When you describe your AI work, do you start with: A) The model B) The feature C) The decision the system makes Comment DECISION. I’ll share how strong candidates structure this in interviews. #productmanagement #aiproductmanager #productleadership #artificialintelligence #careergrowth #techcareers #hiringpanellens

16

3 Comments

Building an LLM - concepts and interview question series Topic 5 – Quantization: PTQ vs. QAT When asked, “How do you optimize a 70B parameter model for inference?” many candidates jump to “Quantization.” But very few can explain how we get from FP16 down to INT8 or INT4 without turning the model into gibberish. The bridge between massive weights and limited GPU VRAM is carefully engineered quantization. So to control the trade-off between perplexity (accuracy) and latency, you must distinguish between two paradigms: Post Training Quantization (PTQ) - This is like a quick win. You take a pre-trained Llama-3 or Mistral checkpoint and calibrate it after training. Mechanics: Uses a small dataset to fix activation ranges. Used - Flows like load_in_8bit or GGUF fall here. (for more see bitsandbytes package documentation) Reason it works: Large models have parameter redundancy. They can "eat" the precision loss with <1% perplexity degradation. Quantization-Aware Training (QAT) – So here you inject fake quantization nodes during a fine-tuning stage. The model “sees” the rounding noise during the forward pass and adapts its weights to survive it. (Fake it till you make it for LLMs :)) Used: Critical for INT4 (and below) or on-device SLMs. Reason it works: At 4-bit precision, naive PTQ often breaks reasoning capabilities. QAT recovers that lost signal. Why this matters for LLMs: A 7B model in FP16 needs ~14GB VRAM. In INT8, it needs ~7GB. This isn't just about saving space; it is about Memory Bandwidth. Since inference is memory-bound, moving 50% less data per token directly translates to higher tokens-per-second. Real-World Example: Meta’s DLRM Recommender systems at scale (like Meta’s) are often bottlenecked by fetching massive Embedding Tables from memory, not by the CPU compute. By applying QAT to these embeddings, they compress Terabyte-scale tables by 4x. This frees up memory bandwidth, allowing the model to run drastically faster in production without losing the nuances about user preferences. My Rule of Thumb: Start with a solid PTQ process. If perplexity degrades beyond tolerance (especially at 4-bits)—pay the compute cost for QAT. The code block below shows how PyTorch simulates INT8 noise during training (the QAT workflow). #PyTorch #LLM #GenerativeAI #MachineLearning #Quantization

23

3 Comments

The dashboard said 94% success. The humans said they were miserable. Who was right? Watch the short clip 👇 This is what Praveen Tomar (Head of Data Automation at Ofgem) discovered after years of chasing numbers: "𝗗𝗼𝗻'𝘁 𝘁𝗿𝘂𝘀𝘁 𝗱𝗮𝘁𝗮 𝗯𝗹𝗶𝗻𝗱𝗹𝘆. 𝗧𝗿𝘂𝘀𝘁 𝘁𝗵𝗲 𝗰𝗼𝗺𝗯𝗶𝗻𝗮𝘁𝗶𝗼𝗻 𝗼𝗳 𝗱𝗮𝘁𝗮 𝗮𝗻𝗱 𝗵𝘂𝗺𝗮𝗻 𝘀𝘁𝗼𝗿𝗶𝗲𝘀." Because every dataset is a collection of human decisions - made under pressure and constraints. → "Low engagement" doesn't capture that the system made jobs harder, not easier → "95% training completion" misses that people clicked through just to finish, learning nothing → "30% faster processing" hides that staff now work unpaid overtime fixing new system errors The data is accurate. The story it tells? Completely wrong. 𝗧𝗵𝗶𝘀 𝗶𝘀 𝘄𝗵𝘆 𝘁𝗵𝗲 𝗺𝗼𝘀𝘁 𝘀𝘂𝗰𝗰𝗲𝘀𝘀𝗳𝘂𝗹 𝘁𝗿𝗮𝗻𝘀𝗳𝗼𝗿𝗺𝗮𝘁𝗶𝗼𝗻𝘀 𝗮𝗿𝗲𝗻'𝘁 𝗺𝗲𝗮𝘀𝘂𝗿𝗲𝗱 𝗯𝘆 𝗱𝗮𝘀𝗵𝗯𝗼𝗮𝗿𝗱𝘀 𝗮𝗹𝗼𝗻𝗲. They're the ones where the human story and the data story actually match. --- 𝗙𝗿𝗼𝗺 𝗧𝗿𝗮𝗻𝘀𝗳𝗼𝗿𝗺𝗮𝘁𝗶𝗼𝗻 𝗥𝗲𝗮𝗹𝗶𝘁𝘆 𝗧𝗩 - 𝗘𝗽𝗶𝘀𝗼𝗱𝗲 𝟯 Unfiltered stories. Ridiculous realities. Yes, as I said - Human. 😇 🎥 Watch the full conversation with Praveen Tomar - including his story about the 54 automations that were technically perfect but missed human purpose entirely: https://lnkd.in/eMC5d4AK What's your experience? When has data told you one story, but reality was completely different? ♻️ Repost if this resonates 👋 Follow me Rujuta Singh for more unfiltered transformation insights #humancenteredtransformation #transformationrealitytv #DigitalTransformation #solvedtogether

174

89 Comments

One of the biggest trade-offs in production RAG systems is retrieval quality vs latency. You can improve results by adding: • reranking • larger candidate sets • richer context But every additional step adds latency. A 2–3 second response may be fine for research assistants. But it can break the experience in customer-facing applications. The best AI architectures are not the most sophisticated ones. They’re the ones that fit within the latency budget of the product and the domain. #AIEngineering #GenAI #RAG #MachineLearning #ArtificialIntelligence

12

3 Comments

NBFCs are up first with AI adoption, are you moving with the times or are planning to be left behind?

10

🌾 𝐈𝐧𝐭𝐫𝐨𝐝𝐮𝐜𝐢𝐧𝐠 𝐒𝐚𝐦𝐚𝐫𝐭𝐡 — 𝐈𝐧𝐝𝐢𝐚’𝐬 𝐈𝐧𝐭𝐞𝐥𝐥𝐢𝐠𝐞𝐧𝐭 𝐐&𝐀 𝐒𝐲𝐬𝐭𝐞𝐦 𝐟𝐨𝐫 𝐀𝐠𝐫𝐢𝐜𝐮𝐥𝐭𝐮𝐫𝐚𝐥 & 𝐂𝐥𝐢𝐦𝐚𝐭𝐞 𝐃𝐚𝐭𝐚 “𝐖𝐞 𝐡𝐚𝐯𝐞 𝐭𝐡𝐞 𝐝𝐚𝐭𝐚. 𝐖𝐞 𝐣𝐮𝐬𝐭 𝐜𝐚𝐧’𝐭 𝐮𝐬𝐞 𝐢𝐭.” That’s what a problem statement told me about India’s agricultural datasets. Thousands of high-quality datasets. Decades of rainfall records. Detailed crop production stats. All public. All free. Yet answering a simple, cross-domain question still takes days of manual Excel work. That frustration led me to build something. 🌿 𝐌𝐞𝐞𝐭 𝐒𝐚𝐦𝐚𝐫𝐭𝐡 (सामर्थ्य = Capability) An AI system that makes India’s open government data actually useful and queryable in plain English. Ask ➜ “𝐂𝐨𝐦𝐩𝐚𝐫𝐞 𝐭𝐡𝐞 𝐚𝐯𝐞𝐫𝐚𝐠𝐞 𝐚𝐧𝐧𝐮𝐚𝐥 𝐫𝐚𝐢𝐧𝐟𝐚𝐥𝐥 𝐢𝐧 𝐌𝐚𝐡𝐚𝐫𝐚𝐬𝐡𝐭𝐫𝐚 𝐚𝐧𝐝 𝐆𝐮𝐣𝐚𝐫𝐚𝐭 𝐟𝐨𝐫 𝐭𝐡𝐞 𝐥𝐚𝐬𝐭 𝟏𝟎 𝐲𝐞𝐚𝐫𝐬 𝐚𝐧𝐝 𝐥𝐢𝐬𝐭 𝐭𝐡𝐞 𝐭𝐨𝐩 𝟑 𝐜𝐫𝐨𝐩𝐬 𝐛𝐲 𝐩𝐫𝐨𝐝𝐮𝐜𝐭𝐢𝐨𝐧 𝐢𝐧 𝐛𝐨𝐭𝐡 𝐬𝐭𝐚𝐭𝐞𝐬.” Get back ➜ 📊 Structured insights 📈 Temporal trends 🔗 Exact dataset citations — all within 5 seconds. 💡 𝐖𝐡𝐲 𝐢𝐭 𝐦𝐚𝐭𝐭𝐞𝐫𝐬 → Policymakers can make faster, data-backed decisions → Researchers spend less time cleaning, more time analyzing → Farmers (via extension workers) get evidence-based recommendations Because good policy deserves good data. ⚙️ 𝐔𝐧𝐝𝐞𝐫 𝐭𝐡𝐞 𝐇𝐨𝐨𝐝 — 𝐓𝐡𝐞 𝐓𝐞𝐜𝐡 𝐁𝐞𝐡𝐢𝐧𝐝 𝐒𝐚𝐦𝐚𝐫𝐭𝐡 🧩 Stack: Flask • LangChain • OpenAI (GPT-3.5-turbo) • FAISS • BM25 • Cross-Encoder Reranking • SQLite • Docker 🧠 Approach: A 5-stage Retrieval-Augmented Generation (RAG) Pipeline built for Indian government data 1️⃣ 𝐐𝐮𝐞𝐫𝐲 𝐄𝐧𝐡𝐚𝐧𝐜𝐞𝐦𝐞𝐧𝐭 → HyDE + entity extraction expand natural language queries 2️⃣ 𝐇𝐲𝐛𝐫𝐢𝐝 𝐑𝐞𝐭𝐫𝐢𝐞𝐯𝐚𝐥 → FAISS (dense) + BM25 (sparse) with Reciprocal Rank Fusion 3️⃣ 𝐑𝐞𝐫𝐚𝐧𝐤𝐢𝐧𝐠 → Cross-Encoder + MMR ensures relevance & diversity 4️⃣ 𝐂𝐨𝐧𝐭𝐞𝐱𝐭 𝐂𝐨𝐦𝐩𝐫𝐞𝐬𝐬𝐢𝐨𝐧 → Dedupes and extracts only key evidence snippets 5️⃣ 𝐀𝐧𝐬𝐰𝐞𝐫 𝐆𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐨𝐧 → GPT synthesizes cited, factual responses — no hallucination 📈 𝐑𝐞𝐬𝐮𝐥𝐭𝐬: ✅ 88% factual accuracy ✅ 98% citation correctness ✅<6s average response time 🧩 𝐊𝐞𝐲 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠𝐬 → The hardest part isn’t the AI — it’s handling messy, inconsistent public data. → Indian datasets often have missing values, mixed units, and inconsistent naming. The best AI is invisible — it just works quietly behind the scenes. 🌍 𝐎𝐩𝐞𝐧 𝐒𝐨𝐮𝐫𝐜𝐞. 𝐅𝐫𝐞𝐞 𝐭𝐨 𝐮𝐬𝐞. 🚀 Experience it live: https://lnkd.in/dJi4yfyx 💻 Source code: https://lnkd.in/dVgScGtZ I’d love to hear your feedback, thoughts, and ideas! Let’s keep pushing the boundaries of what AI can achieve. #ImpactTech #AI4Good #OpenData #PolicyTech #LangChain #MachineLearning #InnovationForBharat #FAISS #RAG #DataForGood #SocialImpact #AI #Innovation #India #buildforbharat

19

2 Comments

Today I explored different chunking strategies for RAG systems, and it’s clear that choosing the right approach directly impacts retrieval quality. Here are the key strategies I understand, along with their trade-offs: 1. Character-based splitting ✔️ Pros: Simple to implement Works consistently across all text types Predictable chunk sizes ❌ Cons: Breaks context mid-sentence Poor semantic understanding Lower retrieval accuracy 2. Recursive character splitting ✔️ Pros: Preserves structure better than fixed splitting Flexible fallback mechanism (paragraph → sentence → word) Balanced chunk quality ❌ Cons: Still not fully semantic Requires tuning (chunk size, overlap) 3. Document-specific splitting ✔️ Pros: Leverages document structure (headings, sections, tables) Better contextual grouping Ideal for PDFs, markdown, structured docs ❌ Cons: Format-dependent Requires custom parsers per document type 4. Semantic chunking ✔️ Pros: Context-aware splitting Higher relevance in retrieval Improves embedding quality ❌ Cons: Computationally expensive Requires embeddings during preprocessing Slower pipeline 5. Agentic / intelligent chunking ✔️ Pros: Dynamic and context-driven Can adapt per use case Potentially best retrieval performance ❌ Cons: Complex to design Higher cost and latency Harder to debug One approach I’m currently exploring as well and that is Unstructured.io. It looks promising for handling multi-format documents (PDF, HTML, DOCX, etc.) and extracting structured elements before chunking. I am Curious about what chunking strategy has worked best in your RAG pipelines and why ? #AI #RAG #Backend #DataEngineering #LLM #SystemDesign

5

Amazon just launched at-home blood testing, India! Sharp move by Amazon—but let’s not pretend they invented at-home diagnostics in India. Players like Thyrocare, 1mg, and others have been scaling this for years. What’s new is the Amazon effect: turning a growing trend into a mass-market habit with logistics, trust, and customer experience. India as a country is not just a launchpad—it’s a stress test, for most businesses it is. If a model works across India 1+, 1, 2, and 3… it’ll likely work anywhere. The question now: Can startups differentiate beyond speed and convenience—maybe on trust? What do you think? Will you use Amazon of existing startup ?

2

3 Comments

What really happens when you click “Delete All Chats” in ChatGPT? With most apps, deleting data is fairly straightforward. It vanishes from a database, assuming good intentions, it’s gone. But AI apps are a different beast. Once a model processes your data, it doesn’t just store it, it may learn from it. Your inputs could shape weights, influence patterns, and subtly steer future outputs. In that world, “delete” doesn’t mean erase. It means unlearn. And unlearning isn’t easy. Machine unlearning - teaching an AI to forget what it has already absorbed is an active area of research. Google even hosted a Kaggle competition on it recently. As AI apps proliferate, this becomes more than a technical quirk. → It’s a governance challenge. → A business differentiator. → A moral obligation. Because the right to be forgotten isn’t just a checkbox. It’s going to define whether AI serves us or just remembers us indefinitely.

9

History has a pattern most people ignore. When two brothers commit to one mission, outsized things tend to happen. These two just put India on the global AI product map. Mukund Jha and Madhav Jha didn’t start Emergent to “democratize coding.” They started it because app building was fundamentally broken. Especially for people building real businesses, not weekend side projects. Most people think building apps is hard because they’re non-technical. That’s a convenient myth. The real problem is the process. Too many tools. Too many handoffs. Too many systems that work fine until real users show up. And that’s exactly when founders can’t afford failure. So instead of hype decks and loud launches, these two brothers did something rare. They went quiet. No buzz. No flashy demos. No “AI will change everything” noise. Just months of grinding through the system, removing friction where it actually breaks. What came out of that effort became Emergent. Now here’s the part that should make you pause 👇 In just 7 months: 5M+ builders 6M+ real apps powered $50M ARR Users across 190 countries That kind of growth doesn’t come from hype. It comes from reliability. Which is why serious investors noticed. Early backing from Google AI Futures Fund, Gossamer Ventures, and SoftBank, followed by a $70M Series B. Not because it sounded exciting. But because it didn’t break when real users arrived. While most AI app builders stop at prototypes, Emergent goes end-to-end: Backend. Frontend. Database. Cloud. Payments. Deployment. Production-ready from day one. So what does this mean for the Indian Tech & IT landscape? This is the real shift 👇 1️⃣ India moves from services to software leverage When app creation becomes this fast and reliable, Indian founders can own products and IP, not just deliver projects. 2️⃣ Pressure on traditional IT models Billing hours and large dev teams won’t disappear overnight, but their dominance will shrink. Outcome-driven software beats manpower-driven software. 3️⃣ Faster startup cycles Founders no longer wait months to validate ideas. They can go from idea → production → users in days, not quarters. 4️⃣ Talent expectations will change The new edge won’t be “knows React or Java.” It’ll be: can you design systems, workflows, and outcomes? 5️⃣ India becomes a serious global builder, not just an executor Tools like this quietly shift India’s role from “cost-efficient tech” → product-first innovation hub. This isn’t just a startup success story. It’s a signal. Software is shifting from: "writing code to describing outcomes" And the teams that win won’t be the loudest. They’ll be the ones whose products don’t fail at scale. 👇 Let’s open this up Do you think tools like Emergent will disrupt Indian IT services or force them to reinvent themselves faster than ever? #IndianTech #IndiaAI #ArtificialIntelligence #AIProducts #StartupIndia #ProductManagement #ProductManager #ProductStrategy #AIInProduct #TechLeadership

11

​Claude Code vs. Indian IT: Why LTIMindtree Hit a 52-Week Low (And Why the Markets Might Be Wrong) ​ The ticker is red, Foreign Institutional Investors (FIIs) are exiting, and the "AI Panic" is real. As the Nifty IT index sheds value, LTIMindtree (LTIM) has corrected nearly 30% from its high of ₹6,430. But as a Wealth Manager, I suggest that we look at what the 1-minute charts are missing. Is this a structural collapse or a classic market overreaction? At FEEDEZ LEARNING, we’ve analyzed the data. Here are the 3 reasons behind the fall and why the "noise" might be creating an opportunity: ​1. The "Exceptional" Hit to Profits In late January, LTIM reported a 30% QoQ drop in net profit. To the casual observer, this looks like a disaster. However, digging into the notes, the company took a one-time charge of ₹590 crore for new government labour codes. Insight: Operational revenue actually hit a record high of ₹10,781 crore. When you strip away the one-offs, the core engine is still firing. ​2. The "AI Panic" (Claude Code & Anthropic) The launch of tools like Claude Code has sparked fears that legacy coding—the bread and butter of Indian IT—will be automated away. Jefferies and other brokerages have downgraded the sector to "Underperform" based on this "deflationary" risk. My Point of View: We’ve seen this movie before with Cloud and Blockchain. While AI will automate routine tasks, it creates a massive $300 billion opportunity in AI integration and "plumbing." LTIM’s recent $155 million mega-deal win proves that enterprise demand is evolving, not disappearing. ​3. Valuation Derating LTIM has always traded at a premium. With growth settling into the high single digits, the market is "right-sizing" that premium. The Silver Lining: With a ROCE of 27.6% and a healthy dividend yield, the stock is entering what I call the "Value Zone" for long-term wealth plans. ​The Verdict: The market is currently pricing in the "Death of IT." I believe we are seeing a structural transition, not an extinction. For those crafting a long-term wealth plan, these are the moments where discipline beats drama. ​Disclaimer: This post is for educational purposes under FEEDEZ LEARNING and does not constitute a direct buy/sell recommendation. Equity investments are subject to market risks. Please consult a SEBI-registered Investment Advisor or Research Analyst before investing. Financial data is based on market observations as of February 2026. ​— Shanker Maharia Chartered Wealth Manager | M.Com, MBA Finance NISM Certified: Investment Advisor (XA, XB) | PMS (XXI-A) | Mutual Funds (VA) ​#LTIMindtree #StockMarketIndia #WealthManagement #FeedezLearning #IndianIT #NiftyIT #InvestingInsights

4

Why Boosting Often Outperforms Bagging on Structured Data ?? In structured datasets like credit scoring, fraud detection, or churn prediction, boosting models often deliver superior performance compared to bagging. Here's why ? 1. Sequential Learning - Boosting builds models iteratively, where each new learner focuses on correcting the errors made by the previous ones. This leads to a significant reduction in bias. 2. Weighted Model Aggregation- Boosting gives more weight to models that perform well, while bagging (like Random Forest) combines models equally. This makes boosting more adaptive. 3. Captures Complex Feature Interactions- Boosting methods like XGBoost or LightGBM are more effective at handling sparse data, high-cardinality features, and subtle non-linear interactions. 4. Built-in Regularization- Boosting frameworks provide shrinkage, subsampling, and tree constraints, making them better suited for preventing overfitting on noisy or imbalanced datasets. 5. Bias-Variance Trade-off- Bagging primarily reduces variance by averaging unstable learners. Boosting reduces both bias and variance, often leading to better generalization. In structured data problems where signal extraction is key, boosting tends to outperform bagging in both accuracy and model interpretability when tuned properly.

8

"Sateesh, I'm stuck at ₹30 LPA. I know I'm worth more. What do I do?" That's the message Amit sent me last week. 8 years of experience. Strong in PySpark. Knows Databricks well. Works at a top MNC. On paper — a solid Data Engineer. But stuck. I asked him one question: "Amit, when was the last time someone asked you to DESIGN a system — not just build one?" Silence. That was the answer. Here's the truth nobody tells you at ₹30 LPA — The market is not paying more for better code. It is paying more for better thinking. The jump from ₹30L → ₹45L is not a technical upgrade. It is a mindset upgrade. So I designed Amit a 20-day personal plan. Not group classes. Not recorded videos. Just us — 20 focused 1-on-1 sessions built around exactly what he is missing: 📐 System Design & Architecture thinking — how to design a platform, not just build a pipeline ❄️ Snowflake — because he was curious, and that curiosity is actually smart market awareness 🤖 Gen AI, LLMs & RAG Pipelines — the intersection of data engineering and AI that is paying ₹50–70 LPA right now ☁️ Multi-cloud Architecture — so no interviewer can corner him on platform choice 📊 FinOps, Governance & CI/CD — the skills that separate architects from engineers 🎯 5 real projects + mock interview — so he walks in with a portfolio, not just a resume In 20 days, Amit will not just know these topics. He will be able to walk into a whiteboard session at Flipkart, Infosys, or a product startup — draw an architecture, defend every decision, and negotiate ₹45 LPA with confidence. If you are reading this and you felt that same silence when I asked Amit my question — This is your message too. Drop me a DM or comment below. Let's build your plan. 🔗 levelupedu.net #DataEngineering #CareerGrowth #DataArchitect #MicrosoftFabric #Snowflake #GenAI #RAG #PySpark #Databricks #LevelUpCloudTrainings #IndianTechCommunity #SalaryGrowth

11

Quantization in Large Language Models (LLMs) As LLMs scale, intelligence is no longer the bottleneck — memory, latency, and cost are. Quantization turns out to be one of the most powerful (and misunderstood) techniques enabling real-world deployment of large models. In this blog, I break down: • Why LLMs are surprisingly quantization-friendly • INT8 vs INT4 vs NF4 — what actually works in practice • How modern architectures absorb low-bit noise • Where quantization fails and how to design around it Quantization is no longer an optimization step — it’s foundational infrastructure for production LLM systems. Link: https://lnkd.in/gqGje75G #LLMs #AIEngineering #Quantization #MachineLearning #GenAI #MLOps

9

2 Comments