Ai Engineering

GPU kernel programming is the hardest skill in AI engineering. Most engineers treat the GPU like a black box. Tokens go in. Latency comes out. The kernel is someone else's problem. Until it isn't. Here's the path fr…

I've never stopped reading books 📚 So here it is: 𝐦𝐲 𝐛𝐨𝐨𝐤𝐬𝐡𝐞𝐥𝐟 (with links!) I love learning by coding. I love reading Substack articles. I love watching technical videos. But sometimes, I need more—more de…

Razorpay asked this question for AI Agent Engineer Role: Your AI agent can spend money using company credit cards to complete tasks. One prompt injection causes it to purchase $50,000 worth of unnecessary services. How …

TCS asked this in L3 round of AI Engineer role: Question: Your RAG data changes every hour. How do you manage versioning without breaking the system? Answer: When data changes frequently, you can't just re-index everyt…

Here is complete roadmap and free resources to master snowflake for data and AI engineering… 1. Snowflake Architecture 2. Snowflake SQL 3. Data Loading & Unloading 4. Semi-Structured Data (JSON, Parquet, XML) 5. Perf…

TCS asked this in L3 round of AI Engineer role: Question: Your RAG data changes every hour. How do you manage versioning without breaking the system? Answer: When data changes frequently, you can't just re-index everyt…

TCS asked this in L3 round of Senior AI Engineer: Answer below 👇 If your RAG chatbot is slow, it’s because you’re treating production like a Jupyter Notebook. Use the following strategies: 1️⃣ Stream responses Show …

Myntra asked this question for MLOps Engineer Role: Here's how to answer it: 1. Track quality metrics continuously, not just at launch Set up automated evaluation pipelines that score a sample of live responses daily. …

𝐀𝐬𝐤𝐞𝐝 𝐛𝐲 𝐈𝐁𝐌 𝐟𝐨𝐫 𝐚𝐧 𝐀𝐈 𝐄𝐧𝐠𝐢𝐧𝐞𝐞𝐫 𝐫𝐨𝐥𝐞. In a RAG system with similarity search, how do you make sure that when someone searches for an exact ID like “Order #1766”, it doesn’t return a similar …

The interview that ended in 30 seconds. Question: "How will you store documents in a Vector DB for a production RAG system?" Candidate's answer: "I'll chunk by paragraph, embed each chunk with OpenAI, and dump everythi…