Lab notes
Research and ideas
Papers and preprints when they exist. Working notes and open questions in the meantime — labeled as ideas, not publications.
Papers
Peer-reviewed work and preprints. PDF and venue go here when a piece is real.
No papers yet
When a paper or preprint is ready, it lands here with title, authors, venue, abstract, and a PDF. Until then this shelf stays empty on purpose.
Ideas
Questions I am actually chewing on while shipping products. Not citations.
From TCP congestion to service backpressure: what networks coursework teaches backend engineers
Networks class gives you vocabulary most backend tutorials skip: congestion, backpressure, routing tables, and what happens when the pipe is full. Those ideas map directly onto queues, rate limits, circuit breakers, and the P99 latency work I did at Cognizant.
What failover actually means when a product depends on six LLMs
Multi-provider routing is usually sold as “if Claude is down, call Gemini.” In a product, failover also means latency budgets, quality drift, cost cliffs, and what the user sees while the chain walks. I want a clearer model for that — not another wrapper SDK.
Fifteen agents vs one planner: when specialization starts to cost more than it saves
Splitting a personal OS into specialized agents (fitness, DSA, job search, deep work) feels clean on a whiteboard. Coordination tax shows up the first week you actually live in it: shared memory, conflicting goals, and the human who still has to be the router.