https://subodhkc.com/blog Articles and field notes on AI systems, governance, decisions, architecture, compliance automation, and production AI. Practical writing from enterprise implementation experience. en Fri, 11 Sep 2026 17:16:46 GMT https://subodhkc.com/blog/where-should-company-use-ai https://subodhkc.com/blog/where-should-company-use-ai A practical framework for identifying AI opportunities worth pursuing, separating business problems from technology ideas, and deciding what should remain human before committing to production. Sun, 16 Aug 2026 00:00:00 GMT AI use case discovery AI use case discovery framework AI use case prioritization AI use case prioritization framework how to identify AI use cases https://subodhkc.com/blog/claude-code-mcp-failed-to-connect-stdio-enoent-timeout https://subodhkc.com/blog/claude-code-mcp-failed-to-connect-stdio-enoent-timeout Claude Code MCP server failed to connect? The error is a status, not a root cause. Diagnose pending approval, ENOENT, stdio startup, stdout corruption, timeouts, the 2026-07-28 protocol change, remote gateway failures, and secure database access boundaries in order. Sat, 08 Aug 2026 18:00:00 GMT claude code mcp failed to connect mcp server failed to connect claude mcp ENOENT MCP stdio server MCP pending approval https://subodhkc.com/blog/modal-fastapi-postgres-ai-voice-pipeline-architecture https://subodhkc.com/blog/modal-fastapi-postgres-ai-voice-pipeline-architecture A production architecture for AI voice agents using Modal, FastAPI and Postgres: multi-tenant routing, three-layer pipeline, latency budgets, graceful degradation, and failure handling. Sat, 08 Aug 2026 15:00:00 GMT modal fastapi voice agent realtime ai reasoning pipeline ai voice platform architecture low latency AI voice agent FastAPI WebSocket voice agent https://subodhkc.com/blog/fastapi-websocket-1006-llm-streaming-disconnect https://subodhkc.com/blog/fastapi-websocket-1006-llm-streaming-disconnect FastAPI WebSocket closing with code 1006 during slow LLM generation? The problem is rarely a single timeout. Diagnose event-loop blocking, Uvicorn ping settings, proxy idle limits, and keepalive architecture correctly. Sat, 08 Aug 2026 14:00:00 GMT fastapi websocket close code 1006 llm prevent fastapi stream disconnect latency fastapi websocket ping timeout asyncio websocket keepalive uvicorn websocket timeout https://subodhkc.com/blog/fix-vllm-cuda-out-of-memory-kv-cache-docker https://subodhkc.com/blog/fix-vllm-cuda-out-of-memory-kv-cache-docker Fix vLLM CUDA OOM startup failures by diagnosing whether the error is a physical CUDA allocation failure or a KV cache capacity error. Two different problems requiring opposite fixes. Sat, 08 Aug 2026 14:00:00 GMT vLLM CUDA out of memory vLLM OOM startup vLLM KV cache memory gpu_memory_utilization vLLM vLLM Docker GPU memory https://subodhkc.com/blog/texas-hb-149-eu-ai-act-engineering-compliance-playbook https://subodhkc.com/blog/texas-hb-149-eu-ai-act-engineering-compliance-playbook Texas HB 149 and the EU AI Act are structurally different laws that create surprisingly similar engineering problems: AI inventory, CI/CD gates, evidence generation, and drift monitoring. Sat, 08 Aug 2026 14:00:00 GMT Texas AI law HB 149 compliance EU AI Act engineering framework NIST AI RMF implementation TRAIGA compliance Texas Responsible Artificial Intelligence Governance Act https://subodhkc.com/blog/12-production-readiness-checks-for-ai-pilots https://subodhkc.com/blog/12-production-readiness-checks-for-ai-pilots 12 production readiness checks for enterprise AI pilots. Each check maps to a specific failure mode, architectural control, and evidence artifact required for audit. Thu, 06 Aug 2026 14:47:09 GMT AI production readiness AI pilot readiness checks enterprise AI deployment AI compliance evidence AI security testing https://subodhkc.com/blog/recover-a-stalled-ai-pilot-in-30-days https://subodhkc.com/blog/recover-a-stalled-ai-pilot-in-30-days Recover a stalled AI pilot in 30 days with structured diagnostic and recovery steps. Sun, 02 Aug 2026 14:58:38 GMT AI pilot recovery stalled AI project AI operations AI deployment AI recovery plan https://subodhkc.com/blog/implementing-immutable-audit-trails-for-soc-2-ai-compliance https://subodhkc.com/blog/implementing-immutable-audit-trails-for-soc-2-ai-compliance This guide outlines practical steps to implement immutable audit trails for SOC 2 compliance, enhancing data integrity and security. Wed, 29 Jul 2026 14:45:22 GMT immutable audit trails SOC 2 compliance AI governance log management data integrity https://subodhkc.com/blog/hidden-seo-risk-ai-assisted-frontend-development https://subodhkc.com/blog/hidden-seo-risk-ai-assisted-frontend-development A technical whitepaper examining how client-only rendering in AI-assisted Next.js and React apps reduces crawlability, delays Google indexing, and obscures public web content with audit checklist and server-first patterns. Wed, 29 Jul 2026 11:00:00 GMT SEO risk AI-assisted development client-side rendering SEO Next.js SEO React SEO crawlability JavaScript SEO https://subodhkc.com/blog/build-a-logging-pipeline-for-eu-ai-act-compliance https://subodhkc.com/blog/build-a-logging-pipeline-for-eu-ai-act-compliance This guide offers practical steps for building a logging pipeline that meets EU AI Act Article 12 compliance requirements. Sat, 25 Jul 2026 14:30:47 GMT EU AI Act logging pipeline AI compliance Article 12 AI governance https://subodhkc.com/blog/what-is-llms-txt https://subodhkc.com/blog/what-is-llms-txt A builder's case for llms.txt, llms-full.txt, and the emerging agentic web with evidence, implementation templates, and a real retrieval case study. Sat, 25 Jul 2026 13:00:00 GMT what is llms.txt llms.txt llms-full.txt AI SEO GEO https://subodhkc.com/blog/haiec-modular-ai-governance-framework https://subodhkc.com/blog/haiec-modular-ai-governance-framework A modular AI governance platform built for EU AI Act, NIST AI RMF, and ISO 42001 compliance. Four core modules: Compliance Engine, Red Audit Kit, Precision Drift Detection, and LegacyShift. Fri, 10 Jul 2026 00:00:00 GMT AI governance AI compliance HAIEC EU AI Act NIST AI RMF https://subodhkc.com/blog/ai-voice-agent-architecture-kestrelvoice https://subodhkc.com/blog/ai-voice-agent-architecture-kestrelvoice How KestrelVoice works under the hood: telephony, adaptive orchestration, RAG, tool execution, degradation modes, security, and the evidence pipeline that makes AI voice agents production-ready. Wed, 08 Jul 2026 00:00:00 GMT AI voice agents AI voice architecture KestrelVoice AI receptionist voice agent production https://subodhkc.com/blog/implementing-rag-row-level-security-for-multi-tenant-ai https://subodhkc.com/blog/implementing-rag-row-level-security-for-multi-tenant-ai Implement RAG row-level security for multi-tenant AI with PostgreSQL RLS policies, tenant isolation architecture, and connection pooling patterns that prevent cross-tenant data leakage. Sun, 05 Jul 2026 00:00:00 GMT RAG row-level security multi-tenant AI PostgreSQL RLS tenant isolation AI RAG security https://subodhkc.com/blog/production-rag-architecture-patterns-for-hybrid-search https://subodhkc.com/blog/production-rag-architecture-patterns-for-hybrid-search Implement hybrid search for production RAG: BM25 lexical plus vector semantic search with reciprocal rank fusion, scoring code examples, and retrieval evaluation metrics. Sun, 05 Jul 2026 00:00:00 GMT hybrid search RAG BM25 vector search reciprocal rank fusion RAG retrieval architecture pgvector hybrid search https://subodhkc.com/blog/seven-layers-ai-compliance-nist-iso-soc2 https://subodhkc.com/blog/seven-layers-ai-compliance-nist-iso-soc2 AI compliance is not a single framework. The seven layers map NIST AI RMF, ISO 42001, and SOC 2 to legal obligations, security testing, architecture, evidence, and governance. Sun, 05 Jul 2026 00:00:00 GMT AI compliance NIST AI RMF ISO 42001 SOC 2 AI security testing