Pradhumn Gupta https://pradhumngupta.com I build production RAG, agentic search, and the systems that make LLMs actually work in the real world. en Why RAG fails in production — and where https://pradhumngupta.com/writing/why-rag-fails-in-production https://pradhumngupta.com/writing/why-rag-fails-in-production When RAG breaks, the model usually isn't the problem — retrieval is, about 73% of the time. Here's the real failure map and what to fix first. Wed, 08 Jul 2026 00:00:00 GMT Why I don't use LangChain in production RAG https://pradhumngupta.com/writing/why-i-dont-use-langchain-in-production https://pradhumngupta.com/writing/why-i-dont-use-langchain-in-production Frameworks are great for a demo. In production, the abstraction you can't see into costs more than it saves. Wed, 08 Jul 2026 00:00:00 GMT When NOT to use RAG (RAG vs fine-tuning vs prompting) https://pradhumngupta.com/writing/when-not-to-use-rag https://pradhumngupta.com/writing/when-not-to-use-rag RAG became the default reflex for every LLM problem. Often it's the wrong tool. A straight decision framework for RAG vs fine-tuning vs just prompting. Wed, 08 Jul 2026 00:00:00 GMT A reference architecture for production RAG https://pradhumngupta.com/writing/production-rag-reference-architecture https://pradhumngupta.com/writing/production-rag-reference-architecture The stages every production RAG system needs — indexing, query-time retrieval, and always-on evaluation — and how to order them. Wed, 08 Jul 2026 00:00:00 GMT Naive retrieve-then-generate isn't enough for production https://pradhumngupta.com/writing/naive-rag-is-not-enough https://pradhumngupta.com/writing/naive-rag-is-not-enough The 5-line RAG baseline is great for a demo and fragile in production. Here's where it breaks and the upgrade path that fixes it. Wed, 08 Jul 2026 00:00:00 GMT The single most useful log line in a RAG system https://pradhumngupta.com/writing/most-useful-log-line-in-rag https://pradhumngupta.com/writing/most-useful-log-line-in-rag If you log one thing in your RAG pipeline, log the final assembled prompt. It's the fastest path from 'bad answer' to root cause. Wed, 08 Jul 2026 00:00:00 GMT Hybrid search for RAG: combining BM25 and dense vectors with RRF https://pradhumngupta.com/writing/hybrid-search-bm25-dense-rrf https://pradhumngupta.com/writing/hybrid-search-bm25-dense-rrf Why you can't average keyword and vector scores — and how Reciprocal Rank Fusion combines them for an 8–15% accuracy gain. Wed, 08 Jul 2026 00:00:00 GMT How to reduce RAG hallucinations: a practical checklist https://pradhumngupta.com/writing/how-to-reduce-rag-hallucinations https://pradhumngupta.com/writing/how-to-reduce-rag-hallucinations Most RAG hallucinations are retrieval bugs in disguise. A concrete checklist to cut them — grounding, retrieval quality, and faithfulness checks. Wed, 08 Jul 2026 00:00:00 GMT How to debug a RAG pipeline that returns wrong answers https://pradhumngupta.com/writing/how-to-debug-a-rag-pipeline https://pradhumngupta.com/writing/how-to-debug-a-rag-pipeline A step-by-step debug order for bad RAG answers — start with the assembled prompt, not the model. Most bugs are retrieval, not generation. Wed, 08 Jul 2026 00:00:00 GMT How to choose chunk size and overlap for RAG https://pradhumngupta.com/writing/how-to-choose-chunk-size-for-rag https://pradhumngupta.com/writing/how-to-choose-chunk-size-for-rag Start at 512 tokens with light overlap, then tune by query type. A practical, benchmark-backed guide to chunk size for retrieval. Wed, 08 Jul 2026 00:00:00 GMT RAG chunking strategies: a practical reference https://pradhumngupta.com/writing/rag-chunking-strategies-reference https://pradhumngupta.com/writing/rag-chunking-strategies-reference The common chunking approaches for retrieval, when each one helps, and the tradeoffs — a quick reference you can come back to. Mon, 06 Jul 2026 00:00:00 GMT