May 15 · 16 min read · This is Part 10 of the AI Engineering with TypeScript series. Prerequisites: Part 2 — MCP Fundamentals · Part 3 — AI Agent · Part 5 — Production MCP Server Stack: Node.js 20+ · TypeScript 5.x · OpenAI
Join discussion
May 12 · 6 min read · Most production AI failures are not model failures. They are retrieval failures. If you want to understand why your Retrieval-Augmented Generation (RAG) system is hallucinating, stop looking at your p
DDigit commented
May 8 · 6 min read · Benchmarking pgvector IVFFlat vs HNSW indexes for production RAG applications I've spent the last three months stress-testing vector indexes in production environments, and the results challenge conventional wisdom about when to use each index type. ...
Join discussionMay 8 · 6 min read · Benchmarking pgvector IVFFlat vs HNSW indexes for production RAG applications I've spent the last three months stress-testing vector indexes in production environments, and the results challenge conventional wisdom about when to use each index type. ...
Join discussionApr 19 · 11 min read · Originally published on my blog Two weeks ago I shipped my news aggregator as a finished Symfony 8 app — fetch feeds, enrich with AI, send alerts, generate digests. Fifty commits later it looks different. The aggregator grew a conversational chat th...
Join discussionApr 7 · 14 min read · Disclaimer: PharmaGuard is a demo project built to explore RAG architecture and pgvector. It is not validated for clinical use. Do not use it to make medical decisions. Always consult a licensed pharm
Join discussion
Mar 19 · 16 min read · The hardest failure mode I’ve seen in enterprise AI systems isn’t hallucination. It’s amnesia. Not “the model wasn’t smart enough.” Not “prompting is hard.” Something more mundane and more expensive: continuity broke, context evaporated, and a human ...
Join discussion
Mar 10 · 6 min read · I sat down to write the pgvector section of this post—the HNSW index DDL, the reranker batching, the metadata filter shapes—and realized I kept reaching for the wrong file. The query I was proud of wasn’t the vector search. It was dedupeRagChunks. Th...
Join discussion
Mar 9 · 8 min read · The $340 invoice showed up on a Tuesday in late January. Not wild money in isolation, but this was for an internal RAG tool serving exactly 11 people — my team at a fintech startup — and when I did the math, that was $30 per user per month for what a...
Join discussion
