Building data systems that see, stream, and scale.
Data engineer working on ML pipelines, real-time data platforms, AI agents, and developer tools.
Computer vision, streaming, and applied AI, with a soft spot for sports analytics.
|
Real-time soccer video analysis. Fine-tuned YOLOv8 finds players, ball, and referee. ByteTrack follows them across frames. GraphSAGE picks out tactical patterns. Served with FastAPI and tracked in MLflow.
|
Benchmarks RF-DETR, SAM2, SigLIP, and YOLOv8 on soccer player detection and segmentation, with reproducible experiments.
|
|
Multi-agent research system for biology. It doesn't just read papers: it pulls PDB structures, runs flexibility analysis, cross-checks the literature, and suggests experiments.
|
🛠️ jod-skillsClaude Code plugins I use every day: a codebase audit that reads only the risky functions, a Hacker News digest, a GitHub trending digest, and an open-source status checker.
|
|
Real-time stock pipeline. Kafka streams ticks for 10 tickers, Airflow orchestrates, and a TimescaleDB star schema stores them. Runs with Docker Compose.
|
Flights, hotels, cars, deals, and billing as 14 microservices that talk over Kafka, run on Kubernetes, and sit behind a React client.
|
More: nobel-dataintelligence (protein stability from molecular vibrations) · contextbox (AI context capture for developers) · voxt (voice-to-clipboard for Linux, in Go) · revolu-idea (multi-agent deep research)
I fix bugs in tools I use. Merged so far:
- lightpanda-io/browser: headless browser for AI. CSS selector spec fixes, Web API behavior, and CDP input events.
- Effect-TS/effect: TOML and YAML parsing edge cases.
In review: Model Context Protocol servers and Python SDK, PostHog, Caddy, DSPy, Modular (Mojo), GitHub MCP Server.
| Languages |     |
| ML & vision |     |
| AI agents |    |
| Data |     |
| Infra & web |      |
Open to data engineering, ML infrastructure, and backend roles, remote or India-based.
