Building a Production-Style AI System: Designing VibeFit from Image Upload to Intelligent RecommendationsLearn how I built an AI recommendation system using LangGraph, Gemini, OpenAI embeddings and pgvector for semantic search.5 min read·blog.ankitapatil.devJul 16
Moving the Brains to the Metal: My Local AI Setup with Gemma 4Stop paying for API tokens. Learn how to run Gemma 4 locally with Ollama and Aider on a constrained hardware setup.3 min read·blog.ankitapatil.devMay 4
Stop Building Digital Graveyards: Solving the AI Latency Tax with Shadow Ingestion. Product Engineer deep-dive: Decouple Gemini AI from the main thread using n8n. Solve data entry friction and latency with Shadow Ingestion. #LLMOps3 min read·blog.ankitapatil.dev1Apr 23
Your AI Stream is Fast, but Your UI is Lagging. Here’s Why.Stop microtasks from starving your UI. Master scheduler.yield() to fix laggy AI streams and keep your INP under 200ms. A Product Engineer's guide.3 min read·blog.ankitapatil.devApr 18
Breaking Down a 68MB React Build: Architecture Fixes That Cut It to 21MBA detailed breakdown of diagnosing a bloated 68MB React production build and the architectural and bundling optimizations that reduced it to 21MB, sig5 min read·blog.ankitapatil.devMar 9
