Tanmay Bohra's Blog https://blogs.tanmaybohra.com Notes on systems, Go, TypeScript, and agentic AI — by Tanmay Bohra. en-IN hello@tanmaybohra.com (Tanmay Bohra) hello@tanmaybohra.com (Tanmay Bohra) Sat, 23 May 2026 15:08:35 GMT 1440 Picking Your LLM Without the Hype https://blogs.tanmaybohra.com/posts/picking-your-llm-without-the-hype/ https://blogs.tanmaybohra.com/posts/picking-your-llm-without-the-hype/ MMLU is saturated noise. SWE-bench Pro is the real signal. Here's how to actually evaluate models — benchmarks that matter, the fine-tune vs RAG vs few-shot question, and what each model tier is actually for. Sat, 23 May 2026 14:45:37 GMT hello@tanmaybohra.com (Tanmay Bohra) LLM evaluationmodel selectionSWE-benchMMLURAG vs fine-tuningfew-shot promptingClaudeGPT-4oGeminiLlamaTTFTcontext windowAI engineering Semantic Caching Is the Cheat Code Nobody Talks About https://blogs.tanmaybohra.com/posts/semantic-caching-llm-cheat-code/ https://blogs.tanmaybohra.com/posts/semantic-caching-llm-cheat-code/ Most LLM teams discover semantic caching too late. Here's how the three-layer architecture works, why the threshold matters more than you think, and what real production hit rates look like. Sat, 23 May 2026 14:45:21 GMT hello@tanmaybohra.com (Tanmay Bohra) semantic cachingLLM infrastructureRAG optimizationGPTCacheRedis LangCachevector searchcosine similarityLLM cost reductionprompt cachingchatbot performance The LLM Deployment Provider Breakdown https://blogs.tanmaybohra.com/posts/llm-deployment-provider-breakdown/ https://blogs.tanmaybohra.com/posts/llm-deployment-provider-breakdown/ Groq vs Bedrock vs Azure OpenAI vs self-hosted vLLM — when each provider actually makes sense, what the compliance tradeoffs look like, and where the $80K/month break-even comes from. Sat, 23 May 2026 14:37:43 GMT hello@tanmaybohra.com (Tanmay Bohra) LLM deploymentAWS BedrockAzure OpenAIGroqvLLMTogether.aiFireworks.aiVertex AIself-hostingHIPAA complianceinference infrastructureML ops Multi-Agent Systems Are Messier Than You Think https://blogs.tanmaybohra.com/posts/multi-agent-systems-production-reality/ https://blogs.tanmaybohra.com/posts/multi-agent-systems-production-reality/ At 95% per-agent reliability, a 10-step sequential chain succeeds 60% of the time. What actually breaks in production multi-agent systems — and what LangGraph, AutoGen, and CrewAI don't tell you. Sat, 23 May 2026 14:33:17 GMT hello@tanmaybohra.com (Tanmay Bohra) AI AgentsLangGraphMulti-AgentProductionLLM Infrastructure The AI Chatbot Scaling Playbook https://blogs.tanmaybohra.com/posts/ai-chatbot-scaling-playbook/ https://blogs.tanmaybohra.com/posts/ai-chatbot-scaling-playbook/ What breaks when your AI chatbot hits 1,000 concurrent users — and how to fix it before it does. Stateless design, semantic caching, gateways, and streaming at scale. Sat, 23 May 2026 10:01:02 GMT hello@tanmaybohra.com (Tanmay Bohra) AI InfrastructureScalingLLM GatewayECSRedis Why Your RAG Falls Apart at Scale https://blogs.tanmaybohra.com/posts/rag-falls-apart-at-scale/ https://blogs.tanmaybohra.com/posts/rag-falls-apart-at-scale/ The real failure rate of RAG in production is 26.4% — not 2.3%. Here's what actually breaks: chunking, retrieval drift, context stuffing, and silent hallucinations. And how to fix it. Sat, 23 May 2026 10:00:30 GMT hello@tanmaybohra.com (Tanmay Bohra) RAGVector DatabaseAI InfrastructureHybrid SearchProduction I Made My Website Bookable by Any AI Agent https://blogs.tanmaybohra.com/posts/mcp-ap2-bookable-by-ai/ https://blogs.tanmaybohra.com/posts/mcp-ap2-bookable-by-ai/ After Google I/O 2026 announced AP2 and Universal Cart, I spent a weekend making tanmaybohra.com callable by Gemini, Claude, or any MCP-compatible AI. Here's exactly what I built — and where the ecosystem actually stands. Wed, 20 May 2026 18:49:40 GMT hello@tanmaybohra.com (Tanmay Bohra) AP2MCPAI AgentsRazorpayFastAPI The Golden Database: How Production AI Agents Are Actually Built https://blogs.tanmaybohra.com/posts/the-golden-database/ https://blogs.tanmaybohra.com/posts/the-golden-database/ Replit tried fine-tuning and abandoned it. The real architecture behind the best AI coding agents is a curated store of high-quality traces, retrieved at inference time. Here's how it works — and when fine-tuning, Agent Skills, or inference-time compute is actually the right answer. Tue, 19 May 2026 15:01:16 GMT hello@tanmaybohra.com (Tanmay Bohra) AISystemsLLM The Cheapest LLM Call Is the One You Never Make https://blogs.tanmaybohra.com/posts/llm-agent-observability-cost/ https://blogs.tanmaybohra.com/posts/llm-agent-observability-cost/ Sourcegraph got a $1M cloud bill three months after launch. Fixie's agent looped 847 times before anyone noticed. These aren't edge cases — they're the default outcome of shipping agents without treating observability and cost as first-class concerns. Tue, 19 May 2026 04:41:54 GMT hello@tanmaybohra.com (Tanmay Bohra) AI AgentsObservabilityCost OptimizationLLMEngineering Inngest for Agentic Systems: Durable Execution Without the Infrastructure Tax https://blogs.tanmaybohra.com/posts/inngest-agentic-architecture/ https://blogs.tanmaybohra.com/posts/inngest-agentic-architecture/ A naive 7-step agent with 3% per-step failure has an 80% end-to-end success rate. Inngest's step-level durability raises that to 99.98% — and eliminates the need to hand-build retry logic, human-in-the-loop state machines, or observability tooling from scratch. Wed, 13 May 2026 17:58:46 GMT hello@tanmaybohra.com (Tanmay Bohra) AITypeScriptSystems A Book Club Changed How I Think About My Work https://blogs.tanmaybohra.com/posts/ssn-book-club/ https://blogs.tanmaybohra.com/posts/ssn-book-club/ Developer life stays close to your own team - until an optional hobby group quietly puts you in the same room as people who have been losing sleep over problems you never knew existed. Sat, 09 May 2026 09:23:57 GMT hello@tanmaybohra.com (Tanmay Bohra) ReflectionPeopleWork LangGraph in Production: What Nobody Tells You https://blogs.tanmaybohra.com/posts/langgraph-production/ https://blogs.tanmaybohra.com/posts/langgraph-production/ State machines, retry logic, and the edge cases that make agentic workflows fall apart at scale. Three months of production LangGraph — the parts the docs skip over. Sat, 09 May 2026 09:22:43 GMT hello@tanmaybohra.com (Tanmay Bohra) AIPythonSystems PostgreSQL Indexing Patterns I Actually Use https://blogs.tanmaybohra.com/posts/postgresql-indexing-patterns/ https://blogs.tanmaybohra.com/posts/postgresql-indexing-patterns/ Partial indexes, covering indexes, and when BRIN beats everything — built from real EXPLAIN ANALYZE output, not toy examples. The strategies that cut query time from 4.8s to 12ms. Sat, 09 May 2026 09:22:42 GMT hello@tanmaybohra.com (Tanmay Bohra) PostgreSQLSystemsSQL Zod in Production: Runtime Type Safety Without the Tax https://blogs.tanmaybohra.com/posts/zod-runtime-validation/ https://blogs.tanmaybohra.com/posts/zod-runtime-validation/ TypeScript's type system vanishes at runtime. Zod bridges that gap — but the naive integration adds overhead you don't need. Here's how we use it at scale without paying the parse tax. Sat, 09 May 2026 09:22:42 GMT hello@tanmaybohra.com (Tanmay Bohra) TypeScriptBackend Go Worker Pools: Handling 10,000 Concurrent Requests on a $20 Server https://blogs.tanmaybohra.com/posts/go-worker-pools/ https://blogs.tanmaybohra.com/posts/go-worker-pools/ The naive goroutine-per-request model looks fine until 50k connections hit simultaneously. Here's the worker pool pattern I actually use in production, and why it cut our memory by 80%. Sat, 09 May 2026 09:22:41 GMT hello@tanmaybohra.com (Tanmay Bohra) GoSystemsConcurrency