Jai Dhyani

After over a decade of experience as an applied AI Software Engineer at Amazon, Meta, and startups and following AI Safety on LessWrong since ~2012, I joined MATS 6.0 where I visited METR and developed evals for RE-Bench, which would go on to be some of the most important evals in the METR AI-task-time-horizons-chart you see everywhere now. I returned to Seattle and started a non-profit, the Seattle Network for AI Alignment Problem Solving. Under SNAPS I raised money for and developed Luthien Proxy, a free, open source AI Control implementation at the API level: https://manifund.org/projects/luthien

Twitter / X

Twitter

LessWrong

jai-d-no-the-other-one

EA Forum

jai

Projects

No grants recorded.