AI engineering and tech-opinions writing only.https://www.arun-mv.com/en-ushttps://www.arun-mv.com/blog/okf-for-a-js-survey-platform/https://www.arun-mv.com/blog/okf-for-a-js-survey-platform/A hands-on guide to Google's Open Knowledge Format on a JS/TS monorepo: bundle layout, a scaffold script, CI wiring, and outcomes after three months.Wed, 09 Sep 2026 00:00:00 GMTAI & Engineeringdeveloper-toolingknowledge-managementagentsmonorepodocumentationhttps://www.arun-mv.com/blog/your-vector-database-is-lying-about-multi-hop/https://www.arun-mv.com/blog/your-vector-database-is-lying-about-multi-hop/Dense retrieval hits a proven geometric ceiling on multi-hop questions and fails silently — similarity scores stay confident even when the evidence is missing.Thu, 03 Sep 2026 00:00:00 GMTAI & Engineeringragretrievalvector-searchmulti-hopevaluationhttps://www.arun-mv.com/blog/your-evals-have-no-tests/https://www.arun-mv.com/blog/your-evals-have-no-tests/A green eval suite isn't evidence your agent works — it's the absence of evidence. Mutation testing shows you exactly how blind your evals really are.Thu, 20 Aug 2026 00:00:00 GMTAI & Engineeringevalsmutation-testingcillm-as-judgereliabilityhttps://www.arun-mv.com/blog/genai-business-impact-case-study/https://www.arun-mv.com/blog/genai-business-impact-case-study/A GenAI project succeeds or fails on whether you can name the business metric it moved — and prove it. The architecture is the price of entry, not the win.Thu, 18 Jun 2026 00:00:00 GMTAI & Engineeringgenairagcase-studybusiness-impactarchitecturehttps://www.arun-mv.com/blog/data-governance-genai/https://www.arun-mv.com/blog/data-governance-genai/In regulated, multi-tenant GenAI, governance isn't a constraint on the build — it's what decides whether you can ship. Design it into the first diagram.Fri, 12 Jun 2026 00:00:00 GMTAI & Engineeringdata-governancesecuritymulti-tenantcomplianceraghttps://www.arun-mv.com/blog/ai-sudo-access-legacy-nodejs/https://www.arun-mv.com/blog/ai-sudo-access-legacy-nodejs/A self-healing feedback loop: tail the logs, feed errors to Claude Code, let it fix, build, and restart — unattended. Here's what it actually fixed.Tue, 09 Jun 2026 00:00:00 GMTAI & Engineeringclaude-codeautomationnodejsexperimentshttps://www.arun-mv.com/blog/hallucination-is-a-system-design-problem/https://www.arun-mv.com/blog/hallucination-is-a-system-design-problem/Ground the model in retrieved evidence, constrain its output, verify its claims, and measure everything. A layered defense against LLM hallucination.Tue, 09 Jun 2026 00:00:00 GMTAI & Engineeringhallucinationragllmopsevaluationhttps://www.arun-mv.com/blog/measuring-rag-quality/https://www.arun-mv.com/blog/measuring-rag-quality/RAG quality by vibes doesn't survive a second engineer. Decompose by stage, build the eval set from real failures, calibrate the LLM judge, and gate CI.Thu, 04 Jun 2026 00:00:00 GMTAI & Engineeringragevaluationragascillm-as-judgehttps://www.arun-mv.com/blog/prompting-techniques-that-work/https://www.arun-mv.com/blog/prompting-techniques-that-work/A research-backed breakdown of the prompting methods — from zero-shot to ReAct — that reliably produce better, smarter, more useful AI output.Wed, 03 Jun 2026 00:00:00 GMTAI & Engineeringpromptingllmchain-of-thoughtreacthttps://www.arun-mv.com/blog/making-ai-production-ready/https://www.arun-mv.com/blog/making-ai-production-ready/Shipping production AI isn't about model benchmarks; it's the reliability, retries, fallbacks, observability, and cost discipline that keep LLM systems alive.Mon, 01 Jun 2026 00:00:00 GMTTech Opinionsreliabilityllmopscostobservabilityhttps://www.arun-mv.com/blog/structuring-agentic-workflows/https://www.arun-mv.com/blog/structuring-agentic-workflows/Five layers — context file, plan gate, atomic tasks, git checkpoints, verification loop — that make AI agents fast and trustworthy instead of chaotic.Mon, 01 Jun 2026 00:00:00 GMTAI & Engineeringclaude-codecursoragentsworkflowhttps://www.arun-mv.com/blog/graphiti-episodic-memory/https://www.arun-mv.com/blog/graphiti-episodic-memory/Episodic, bi-temporal memory with Graphiti lets an AI agent answer not just what's true now, but what was true when, without re-indexing the whole world.Wed, 20 May 2026 00:00:00 GMTAI & Engineeringknowledge-graphsgraphitimemoryagentshttps://www.arun-mv.com/blog/multilingual-rag-code/https://www.arun-mv.com/blog/multilingual-rag-code/Code is many languages plus an identifier dialect tokenizers shred. Use code-specialized embeddings, identifier-aware retrieval, and structure-aware chunking.Sat, 16 May 2026 00:00:00 GMTAI & Engineeringragembeddingsretrievalcode-searchtokenizationhttps://www.arun-mv.com/blog/agentic-workflows-guardrails/https://www.arun-mv.com/blog/agentic-workflows-guardrails/Reliable agents come from control, not capability. Cap turns and time, push predictable steps into a state machine, and keep a human on the irreversible ones.Thu, 07 May 2026 00:00:00 GMTAI & Engineeringagentsreliabilityguardrailsstate-machinellmopshttps://www.arun-mv.com/blog/hello-world/https://www.arun-mv.com/blog/hello-world/Vector-only RAG returns flat, context-poor chunks ranked by similarity. Knowledge graphs model entities and relationships to traverse why things connect.Sat, 02 May 2026 00:00:00 GMTAI & Engineeringragknowledge-graphsneo4jllmhttps://www.arun-mv.com/blog/choosing-a-vector-database/https://www.arun-mv.com/blog/choosing-a-vector-database/Every vector DB benchmark was run on someone else's data. Benchmark your own vector count, dimensions, and — above all — your real filtering pattern.Fri, 24 Apr 2026 00:00:00 GMTAI & Engineeringvector-databasepgvectorragretrievalbenchmarkshttps://www.arun-mv.com/blog/poc-to-production-rag/https://www.arun-mv.com/blog/poc-to-production-rag/A RAG demo proves the happy path exists. Production is everything else — tracing, drift, evals, and learning to say 'I don't have that part.'Thu, 09 Apr 2026 00:00:00 GMTAI & Engineeringragproductionobservabilityevaluationllmopshttps://www.arun-mv.com/blog/reducing-llm-hallucination/https://www.arun-mv.com/blog/reducing-llm-hallucination/Most LLM hallucination is a retrieval failure in disguise. Fix the context first, force citations, and give the model a sanctioned way to say 'I don't know.'Mon, 30 Mar 2026 00:00:00 GMTAI & Engineeringhallucinationragretrievalevaluationllmopshttps://www.arun-mv.com/blog/cutting-llm-inference-cost/https://www.arun-mv.com/blog/cutting-llm-inference-cost/Inference cost optimization is a measurement problem in disguise. Fix the quality metric first, then trim context, route models, and cache the stable prefix.Wed, 18 Mar 2026 00:00:00 GMTAI & Engineeringllmopscostragevaluationcachinghttps://www.arun-mv.com/blog/rag-vs-finetuning-vs-long-context/https://www.arun-mv.com/blog/rag-vs-finetuning-vs-long-context/'Should we fine-tune?' is usually the wrong first question. Ask instead: is the gap knowledge or behavior, and how often does the answer change?Thu, 05 Mar 2026 00:00:00 GMTAI & Engineeringragfine-tuninglong-contextarchitecturellmops