available for work
I build systems that
think, talk, and scale.
Real-time AI pipelines, distributed voice infra, and production deployments on AWS. I care more about latency budgets and failure modes than demo polish.
things i've built
- ◎Real-time Voice AI system (production)
- ◎Agent-based backend (FastAPI + ML)
- ◎AWS ECS deployments (Dockerized)
- ◎Raspberry Pi home automation
- ◎Telemedicine for pets (early)
systems
Voice AI System
productionStreaming STT → LLM → TTS
WebSocket orchestration
AWS ECS deployment
Parallel chunk streaming reduced latency from ~1.8s to ~600–800ms. It handles drop / retry / reconnect without derailing the call.
PythonFastAPIWebSocketsAWS ECSDocker
Agentic Backend
FastAPI + ML decisions
Docker + CI/CD
Structured decision pipelines with clear API boundaries. The agent layer stays swappable and the service layer stays testable.
FastAPIDockerCI/CDPython
now
Building real-time AI pipelines, exploring agent workflows, and experimenting with hardware automation.
how i think
- Systems > tools — understanding the whole beats knowing every library
- Latency is a product decision — every millisecond is felt by the user
- Build → break → fix — the fastest way to understand production behavior
- Failure modes first — if you haven't thought about how it breaks, it's not done
writing
topics
awsinfralatencyllmpipelinesrealtimestreamingvoice-ai