How I wired Qdrant, Jina embeddings, and Gemini together — and the retrieval and prompting tricks that kept answers grounded in the source articles.

Token-by-token streaming makes a chatbot feel alive. A practical look at backpressure, reconnection, and cleaning up half-finished streams.

Pagination, indexing, and the small schema decisions that keep a Mongo-backed API fast as data grows. Lessons from my AI note-taking app.

Having shipped both in production, here's how I actually decide — team shape, app complexity, and how much structure you want handed to you.

The mental model transfers, the vocabulary doesn't. Notes on the friction of moving between two clouds without losing a weekend.

The most maintainable code I've written was also the most boring. On writing for the next engineer — who is usually future me.