I work across ML and systems — fine-tuning pipelines, multi-agent orchestration, concurrent Go services, embedding infrastructure. Most of what I build starts from a real constraint: OT networks that can't run heavy security agents, dev workflows that break down once you add more than one AI in the loop, pipelines that need to be fast and correct at the same time.
If you're hiring or evaluating: I tend to own problems end-to-end and push code that works in production, not just demos. The three projects below are the ones I'm most serious about — each started from a specific gap I wanted to close, not a trend I wanted to chase.
Anomaly detection for OT/ICS networks — the kind of infrastructure (PLCs, SCADA, industrial protocols) that traditional security tooling can't touch without breaking things. Runs lightweight ML models over network traffic, surfaces deviations in real time, and feeds a live alerting dashboard. Built because most OT environments are air-gapped, resource-constrained, and one misconfigured firewall rule away from a bad day.
Multi-agent coding platform where multiple engineers and AI agents work in the same session simultaneously — each agent scoped to a task, with its own tool access and context. The core problem: today's AI dev tools are single-player and single-agent. Overmind is what happens when you take that constraint away.
-FF6F00?style=flat){kind=icon}
Multiplayer prompt collaboration layer for Claude Code. A host starts a party and participants join via invite code — each submitting prompts that a Gemini deduplication agent merges into a single consolidated multi-step prompt, executed by the host's Claude Code CLI and streamed back to everyone in real time. Solves the coordination problem when multiple engineers want to steer the same AI session.
