I build product systems end to end: the interface people use, the APIs and background jobs behind it, and the operational details that keep it fast and affordable. Led a ~6-person cross-functional team; launched startup products that reached production, real users, and revenue. AI-native workflow: I ship daily with Claude Code and agent pipelines, and build LLM features as an engineer — cost tracking, retries, and provider abstraction included.
| Project | What it demonstrates | CI | |
|---|---|---|---|
| 💬 | RelayDesk | Realtime support inbox — WebSocket gateway, optimistic sends with idempotent retries, presence, RBAC, audit trail | |
| 🤖 | FlowPilot | LLM workflow engine — provider abstraction, retries with jittered backoff, token accounting, per-run cost budgets | |
| 📊 | CostLens | Cloud-cost intelligence dashboard — typed domain logic, anomaly detection, Docker, tests, polished responsive UI | |
| 🎬 | StreamForge · in progress | Media pipeline — processing jobs, CDN-aware delivery, performance instrumentation |
Portfolio projects are independent demos with fictional data — they show how I work without exposing NDA-protected products or customers.
RelayDesk — realtime inbox with presence, optimistic sends, and audit trail |
FlowPilot — LLM pipelines with cost budgets and per-step token accounting |
| AWS cost reduction (regions, Lambda, CloudFront) |
Twitter/X API cost reduction |
video player initial load |
LLM token-spend reduction* |

