Investing in Chai Discovery
Today, the founding team at Chai Discovery is open sourcing their cutting edge foundation model Chai-1. We're also announcing that Dimension has co-led Chai Discovery's founding round, alongside friends at Thrive and OpenAI.
Chai Discovery is founded by a small, extremely ambitious, technically polymathic team of ML, life science, and computer science practitioners. My cofounders, Nan and Adam, and I, have all closely tracked the work of Joshua Meier - a Chai Discovery founder, since his days at Meta and OpenAI in 2019. Some fun inside baseball: Joshua was amongst the first people on the planet who knew of Dimension when we first founded the firm in 2022 - as Joshua, a natively bilingual ML and life science builder, was archetypal of the ilk of founder to whom we knew existing investment firms were often forcing dichotomous "tech or biotech" faustian bargains.
As we’ve spent time with the entire founding team we’ve seen just how much a small but extremely capable team can accomplish in extremely short order. In between our conversations ranging the limits of ML and mathematical logic, and hosting MLxBio dinners in SF and the UK, the Chai Discovery founders (many of whom were prior ML leads at today's leading "techbio" cos) have rapidly developed Chai-1.
Stunningly, in just a few months, Chai Discovery has brought Chai-1 to parity or supremacy over the existing, long standing and financially well coffered incumbents in industry. As part of today's launch this "rebel alliance" is open sourcing their work. I'm proud that the team is both contributing to open science, but also not resting on the laurels of a snapshot of any single model at any moment in time. Given the continued acceleration and rapid maturation of fundamental ML applied to the current phase shift of "transforming biology from science into engineering" we're steadfast believers that the long term winning strategy necessitates transparency and open experimentation. For today’s biotech practitioners, there’s inevitably translatable drug discovery value, here for free, at your finger tips.
I encourage everyone to sift through Chai-1’s repo, and directly experiment on the hosted webapp. I personally am most excited about applications of multi-modal fine tuning (eg with mass spectra) and prompting with physics informed docking - though I suspect the broader the community will be critical in imagining entirely new applications as well as finding bugs in the current release.
Finally, Chai Discovery is growing their team. If you're world class in the technical worlds of ML and life sciences, I highly encourage you to reach out to the team directly.