Jonathan Lebensold
I build tools that make AI systems trustworthy.
As CEO of Jetty, I’m building infrastructure for accountability, traceability, and reproducibility in AI deployment. In a world racing to ship AI, we help teams actually know what their models are doing – and prove it.
What I’m Working On
I currently serve as CEO of Jetty and am part of the Mila Startup Program, collaborating with MLCommons and Mila on open-source tools for model provenance and AI governance.
I write about the lessons learned behind shipping reliable AI systems on my Substack, Ground Truth.
Background
My career has moved between building and research. I co-founded a software consultancy, shipped products for startups and enterprises, and wrote the React Native Cookbook for O’Reilly.
Then I went deep on the science side – completing a PhD at McGill University and Mila under the supervision of Borja Balle and Doina Precup. I worked on privacy-preserving machine learning at Meta AI and AI reasoning at Reliant AI.
Research Interests
My research focuses on making machine learning systems safe and accountable in practice:
- Differential privacy and privacy-preserving ML – training models that protect individual data while remaining useful
- Generative models – understanding the capabilities and failure modes of modern generative systems
- AI governance and model provenance – building practical tools for tracking what models do, how they were trained, and whether they can be trusted
- AI evaluation and reliability – moving beyond “it works on the benchmark” to real-world accountability
You can find my publications on Google Scholar.
Get in Touch
If you’re deploying AI in healthcare, finance, or any domain where “trust me, it works” isn’t good enough, I’d love to talk.
- Email: jon(at)lebensold(dot)ca
- Substack (Ground Truth)
- Google Scholar