What happens to the messages your AI just couldn’t handle? They don’t disappear they go somewhere very specific.Continue reading on Medium »
Two similarity score formulas. One silent assumption. A ~270× discrepancy waiting to bite you.If you’ve read the Databricks Vector Search documentation, you’ve probably come across two similarity score formulas:Cosine-form: score = 1 / (3 − 2·cosθ)Eucl…
Most AI systems don’t fail because of the model. They fail because of how the knowledge is written.Continue reading on Medium »
Hi everyone. With all the updates in the LLM stack over the past year, I decided to put together a practical list of RAG approaches that are actually useful in production or at least worth understanding if you are building LLM-based products.This is ba…
Breaking down GenAI into simple concepts and real implementationsContinue reading on Medium »
Most RAG tutorials say:Continue reading on Medium »
A two-stage pipeline that looks at your sick plant, reads the textbooks, and gives you an evidence-backed diagnosis. No cloud, no API keys…Continue reading on Medium »
“Your AI agent can follow instructions. But can it write new ones?” — Google Developers Blog, April 2026I’ve been building AI agents for a while now. And if there’s one thing that every developer hits eventually, it’s the context window wall. You start…
By Amrita — The Product ScientistIt was 2 AM when I realized we had a problem.We’d spent three months building a fertility intelligence engine — a system designed to synthesize clinical biomarkers, Ayurvedic phenotyping signals, and lifestyle data into…
Last week we combined vector and keyword search to improve retrieval quality.Continue reading on Operations Research Bit »