Most AI engineers can't sell. Most salespeople can't build. One person, both halves — and every number below is public and checkable.
Design-faithful builds on modern stacks, delivered for paying clients. Every screenshot links to the real, live site.

Whether you need the AI shipped or the revenue closed — or someone who does both — I take on a small number of client projects at a time.
Independently verifiable — every package below is public under my own accounts, not a company's.
The layer that lets LLMs act on real systems — wallets, CLIs, browsers, enterprise APIs (Hedera, Solana, Sui, GOAT, Shopify, IBM watsonx, the full Apple suite). 86 on npm, 46 Rust crates, all under my own handles.
6 models and a live demo on Hugging Face, plus 16 authored agent skills — all public and browsable.
Not client work — my own bets, in the open, each at a different stage of finding its place.


LLM inference on decentralized GPU at 180ms p50, vs. centralized cloud. Companion video pipeline: 16× faster encode at VMAF 95.8.
I don't just ship product — I've owned a quota and hit it at scale.
Led enterprise and emerging-venture onboarding end-to-end across the US–LATAM corridor for a top-tier exchange — priced listing packages, ran the close, and drove ecosystem business development at the industry's largest conferences.
Whether you need the AI shipped or the revenue closed — or someone who does both — I take on a small number of client projects at a time.