Work
Production AI, shipped.
Six products I designed, coded, and deployed end-to-end. Three are deep-dive case studies - the kind of system you can take to an interview and defend. Three more are infrastructure and platform pieces underneath.
Deep case studies
01 · Meeting intelligence
Beaver AI
Bot-joined meetings, transcripts, AI summaries and action items, semantic search across history. Eight integrations. Subscription billing with webhook idempotency.
Read case study →
02 · Voice AI
VoiceDesk
An always-on AI receptionist on phone or web. Full-duplex WebRTC, NAT traversal, pre-loaded knowledge that beats RAG on latency. 1–3 second responses.
Read case study →
03 · Autonomous SDR
Brimley
Private-cloud sales agent: sources prospects, writes per-lead sequences, sends from the customer's own mailbox, classifies replies. Real state machine, ETA-based dispatch, cost-tracked AI.
Read case study →
Platform & infrastructure
04 · Marketplace
FluxAgents
Enterprise AI agent marketplace serving 9 industries and 14 departments. Custom agent builder, 1,000+ integrations, per-hour billing, on-prem deployment.
Read more →
05 · Image generation
Flux Image Engine
Production image-generation server built from scratch on open-weights image models. Custom queuing and batching. Has served hundreds of thousands of images at scale.
Read more →
06 · Bare metal
AI Inference Stack
Full vertical ownership from hardware to API: bare-metal GPU builds, CUDA, model serving, vLLM, load balancing, monitoring. The plumbing under every product above.
Read more →
Want to talk about one of these?
Happy to walk through any of them in detail - architecture, decisions, the parts that broke at 3am. Drop me a line.