All posts by Alpesh Kumar
Design an Uber Demand Prediction and Driver Repositioning System
Interview question:Uber predicts that ride demand in a particular area will increase significantly in the next 20 minutes. How would you design an AI system that predicts demand and proactively moves drivers to that area? This is a classic marketplace + machine-learning system design question. The interviewer is testing whether you can go beyond “build…
Read More
AI Agent Frameworks Compared: Vercel AI SDK, LangGraph, CrewAI, AutoGen and LangChain
A practical, source-led comparison of five AI frameworks. Choose from your application architecture, streaming needs, and agent workflow rather than a universal ranking.
AI Watermarks Are Not a Silver Bullet: What Provenance Signals Can and Cannot Prove
C2PA metadata and embedded signals can offer helpful evidence about an image’s origin. They cannot, on their own, prove that an image is accurate, legally owned, or shown in the right context.
The UUID Primary Key Problem That Shows Up at 200 Million Rows
A primary-key lookup can still become expensive when random UUIDs, cache pressure, and table layout collide. Here is how to diagnose the evidence before changing your schema.
Your First Personal AI Agent: A Focused Weekend Build
Build a small, reviewable AI workflow in a couple of hours by giving it one narrow job, clear boundaries, and a human approval step.
Build an AI Agent Evaluation Flywheel That Improves Prompts
A five-stage evaluation flywheel for turning AI agent traces into evidence-backed prompt improvements, without treating final answers as the whole story.
Why Final-Answer Evals Leave AI Agent Failures Invisible
A correct final response can hide failed tool calls, broken constraints, and bad state changes. Learn why AI agent evaluation must score the full trajectory.
Vector Quantisation: How Code Books Compress Similar Data
Vector quantisation replaces each group of values with the index of a representative vector. Learn how code books, k-means, distortion, and trade-offs fit together.
Contextual Retrieval: Anthropic’s Approach to Reducing RAG Retrieval Failures
Anthropic’s Contextual Retrieval adds document-level meaning to RAG chunks before embedding and BM25 indexing. Here is how it works, what the 67% result means, and when to use it.