Instructor
Michael Zeligs
Frontier AI Trainer & Model-Evaluation Specialist, CEO CXO
About me
I train and evaluate frontier AI models on the hardest problems we can give them — the executive-decision, multimodal, and creative tasks where models still fall short of expert humans. My work is currently performed for a leading U.S. frontier AI lab (engagement confidential) through Handshake AI's MOVE program and Mercor's elite frontier-training cohort.
Contact Me: mzeligs@alumni.stanford.edu
What I bring is a trifecta most evaluators don't have:
• Real executive judgment. I spent 15+ years as a founder/CEO and go-to-market leader — owning P&L, raising and supporting institutional rounds, and driving $120M+ in directly attributed client funding within a $1B+ portfolio. When I author a "golden" reference solution for a CEO- or CMO-level task, it reflects work I've actually done.
• Professional design and brand craft. I'm a working UX/UI and visual designer (Figma, full Adobe Creative Suite, design systems, editorial and presentation design). On AI projects I apply pixel-level judgment to image generation and AI-generated decks.
• Hands-on engineering. Stanford CCRMA trained me in digital signal processing and computational systems; I write Python, C++, and JavaScript/React, and I reason fluently about model behavior, evaluation harnesses, and failure modes.
Right now I design structured task specifications that compress 80-hour executive workflows into single, machine-evaluable problems — building the live-system, multi-step "world-building" tasks meant to break frontier agents, then diagnosing chain-of-thought failures and hand-authoring the correct solutions. I also evaluate generative music as a music-criticism engineer, scoring naturalness, instruction-following, and human preference against a full library of acoustic components.
Earlier in my career I led product marketing, brand, and communications for AI, clean-energy, consumer-electronics, and hardware companies, and I founded Start Motion Media, a creative and communications studio serving clients from the Smithsonian and Discovery Channel to Amazon Fire TV.
I'm open to full-time and contract roles in frontier AI training and evaluation, UX/UI and creative-systems design, and music/audio model work.