S

I lead the AI team at IMPOSSIBLE, a crypto incubator and launchpad, after six years in crypto research at Nethermind, Lemniscap, and DegenDAO, I now build AI tools for crypto and beyond. What I do now: Run a multidisciplinary team shipping AI products for internal teams, retail users, and external clients. We built Model Router, an orchestration layer that picks the right model and provider for each request based on task complexity, quality, latency, availability, and cost. We built an Agent Hosting Platform for deploying and scaling AI agents. I own product strategy across the portfolio and work directly with clients from discovery through post-launch, on products like due-diligence automation and research tooling. Where I come from: I've been in crypto since 2015 and working in it full-time since 2021. I designed tokenomics at DegenDAO and got ranked #4 on DegenScore v1. At Nethermind I did DeFi research for external clients and designed the process Lido uses to maintain its validator set. As Research Lead at Lemniscap I set the research agenda across DeFi, infrastructure, and protocol economics; my work with Gattaca supported Titan Builder's growth from roughly 3% to 30% of Ethereum block building. I've designed token models for Hundred Finance, Mithraeum, and a veToken protocol. Why both together: Crypto taught me to work with adversarial data and incentive-driven systems, which turns out to be good preparation for shipping AI that has to hold up in production. When my team builds a due-diligence agent, I know what good output looks like because I produced it by hand for years. And when a crypto team asks what AI can actually do for them, I can answer from both sides of the table. Technical environment: OpenAI GPT, Anthropic Claude, Google Gemini, AI agents, RAG and vector retrieval, model routing, LLMOps, prompt engineering, model evaluation, Python, SQL. Crypto: DeFi protocols, tokenomics and mechanism design, protocol economics, on-chain analytics (Dune, Nansen).