Fullstack engineer. Early-stage products, zero to one. Smart contracts, L1 infra, and increasingly AI
I work best early: the zero-to-one stage, before the architecture is settled and while the requirements are still moving. Most of what I've built has been the first version of something.
Somnia — Head of Engineering for an L1 blockchain. Took it from devnet through testnet to mainnet and TGE with a team of five. Built the staking and validator-committee protocol that holds 20%+ of total token supply (Hacken audit), and a Sybil-resistant incentive platform that verified and rewarded 95,000+ contributors.
Charms — the Web3 side of a consumer app where AI companions have their own on-chain economies: per-character token launches on Uniswap V4 (through Clanker), gasless one-tap transactions (Privy, EIP-7702, ERC-7579), and cross-chain swaps across Base, Solana and BNB.
Earlier — I joined crypto in NFT Summer 2021: DAO governance and treasury contracts at Improbable, then Gokaden through the startup incubator program at Station F, and a year freelancing on NFT project contracts.
AI product engineering: agents that touch real systems, LLM features that have to survive production, and the evaluation work that decides whether either one ships.
Smart contract work turns out to be strange but useful preparation. You learn to assume inputs are adversarial, to scope exactly what a system is authorized to do, to keep simulation separate from execution, and to never let an unreliable component own an invariant. Those questions arrive the moment a model can take actions instead of just generating text.
Deep — Solidity, Foundry, Hardhat, upgradeable patterns, account abstraction, DeFi, NFTs Daily — TypeScript, Next.js, React, Node, NestJS, Go, Postgres, Docker, AWS/GCP Learning — agent frameworks, evals, LLM tooling
Writing on my blog
