Cibi is
I work on scalable oversight. How do you evaluate an AI’s work once it’s better than the people checking it?
Most alignment techniques need an overseer who can tell good work from bad. That gets harder as models improve, and it gets hard first in alignment research itself, as labs hand more of it to AI agents. I want to test oversight methods now, on tasks where humans can still check the answer, and measure how often they get it right.
Before this, I spent eight years building new products as an engineer, product lead and founder. In 2026 I turned down product leadership roles to work on safety full-time. Since then I’ve done BlueDot’s AGI Strategy and Technical AI Safety courses and had 300+ conversations with people across the field.
Now
- Atlas
- I founded and run Atlas Research, an early-stage technical AI safety org. The agenda is scalable oversight, automated alignment research, and the evaluation tools to tell whether an oversight method works. I’m looking for a research co-founder to lead the technical work.
- Crux
- Atlas’s first concept. A second model tries to break an AI’s research results before a researcher relies on them. I built a scripted demo at BlueDot’s AI for Epistemics hackathon, with Julian Quick, an incoming alignment scientist at AE Studio, as design partner. Read the postSee the deck ↗
- Open to
- Alignment research fellowships and roles, especially in oversight. Also a research co-founder for Atlas, collaborators and funders. Email me.
Experience
- 2026
- Atlas Research Founder & CEO Technical AI safety org, early stage. Prototyped Crux, a reviewer for AI-generated alignment research.
- 2024–26
- Nirvana Founder & Head of Product Mindfulness tools for people in tech. Recruited a technical co-founder, got 100+ sign-ups including people at OpenAI, Meta and Coinbase, and enterprise pilot offers.
- 2021–24
- BLD.ai Lead Product Manager · Head of Venture Studio Promoted twice, to Lead PM. Grew an AI meeting assistant to 70K monthly users. Led product for a performance and pay platform used by Uber, Coinbase and Two Sigma. Ran a venture studio that launched seven products. One was acquired and one raised $1.2M.
- 2020
- Lifion by ADP Technical Product Manager, Mobile Led 12 engineers and 2 designers to launch a low-code platform, from zero to seven-figure ARR across six enterprise customers.
- 2019
- Faire Software Engineer, Growth Joined as the third iOS engineer. Built search and checkout features.
- 2017–19
- Loblaw Digital, TribalScale, Inkling Software Engineer Mobile engineer. Worked on Carnival’s iOS app and built CBS Radio’s Roku app.
- Education
- University of Waterloo Computer Science, Business Option