About
// the short version
I’m a software engineer who likes the unglamorous middle of the stack — the places where data, systems, and AI start fighting each other. I started out on Java microservices at Cognizant, tuning Spring Boot services and watching Kafka graphs settle. From there I crossed into full-stack work, then into AI systems engineering at Buildway, where I build agentic retrieval pipelines, MCP tool servers, and the boring backbone (billing, integrations, streaming) that keeps the interesting parts useful.
What I care about is shipping the unglamorous 90% well so the interesting 10% can exist. I read more than I write, take long walks to debug things, and keep a running theory that almost every AI bug is actually a state-management bug in disguise.
Agentic RAG at Buildway — MCP tool servers on top of a 2.3M-record legal corpus, E2B-isolated execution, Weaviate retrieval.
Going deeper on Go for systems work, distributed tracing with OpenTelemetry, and the internals of vector indexes.
Designing Data-Intensive Applications (re-read), and whatever LLM systems papers cross my feed.
A small terminal-first knowledge base for personal notes — embeddings, MCP, the whole machinery, on one binary.
- 01boring infra beats clever infraIf a system needs a clever explanation to stay up, it won't.
- 02read the actual docsNot the ones you imagine they say. Most AI bugs are docs you skipped.
- 03ship the unglamorous 90%Auth, billing, retries, observability. The interesting 10% only exists if the rest works.
- 04if it's hard to test, it's hard to keepTestability is a design signal, not a chore.