Yariv Barsheshat

AI safety researcher · software engineer · former mathematician

I trained as a mathematician and taught calculus for seven years before moving into industry to build real-time computer vision systems and applied LLM products. I'm now an independent AI safety researcher, working with Robert Graham on a project supported by Coefficient Giving. Our first paper is currently under peer review at NeurIPS and ICML 2026 — more soon.

Experience

Apr 2026 – present: Independent AI safety researcher Coefficient Giving (grantee)

Working with Robert Graham on a research project in AI safety. First paper under peer review at NeurIPS and ICML 2026.
Jan – Apr 2026: AI full-stack engineer Explorai

Contributed across the full stack to Estimai, a SaaS that lets construction estimators generate bids by ingesting architectural plans and tender documents with AI-driven analysis. Built the proof-of-concept for vectorial column detection across multi-page floor plans.
Aug 2022 – Jan 2026: Senior R&D programmer mtl.ai

Trained and deployed real-time computer vision models for live soccer broadcasts (UEFA Euro 2024, Champions League). Built a unified API layer over frontier LLMs powering Tavendi.
Jun 2015 – Jul 2022: Mathematics professor Dawson College

Taught calculus, linear algebra, statistics, and probability; developed course materials adopted department-wide. Executive on the Teachers' Union and the Science Program Committee.

Education

2013 – 2015: MSc Mathematics McGill University

Thesis: Entropy in Dynamical Systems. Shortlisted for top master's thesis in mathematics.
2009 – 2012: BSc Joint Honours, Mathematics & Physics McGill University

Dean's Honour List, GPA 3.91.