As we build the next generation of biomolecular OpenFold models, we are looking to increase our staffing on both the scientific and engineering fronts. I thought it'd be worthwhile to write down our philosophy and what we're offering, as well as what we're looking for. If this resonates with you, please apply at the link below. What we're offering: True open source: we are contractually obligated by our funding streams to open source everything we do, including our training code infrastructure and training data, which no other group in the space has done. All the results we publish are fully reproducible with no missing "training enhancements" that keep the recipes hidden. We will not become a startup and we will not pivot to an API model. We are doing this exclusively as a public good for the public commons. Of course, commercial companies can and are encouraged to build on top of our models, but you will be contributing to an ecosystem guaranteed to remain open source. Long horizon research: because we are not trying to turn a profit or satisfy investment trends, we can focus on hard problems that commercial labs are disincentivized to tackle. Our research vision spans years, not months, even while we release updates on a quarterly basis. This translates to problems spanning conformational ensembles, novel regions of chemical space, antibodies, nucleic acids, IDRs, macromolecular machines, and more. Reach for the frontier: the open source ecosystem currently lags the industrial frontier by a few years. We aim to eliminate this gap. The two big deltas are engineering expertise and compute resources. We're doing alright on the latter, but it is the former where we need to grow. To make this possible we offer salaries and compute resources competitive with industry. We are looking primarily for types of people, or the intersection of the two! Machine learning scientists and engineers: people who have built and trained models at scale, whose command of the literature extends to the latent developments in efficient diffusion, optimization, recurrence, and reasoning. Bioinformaticians and computational biologists: people who can assemble large and complex heterogeneous datasets, who write efficient data pipelines, and who deeply understand the structural biology domain. If this is of interest to your network, please reshare. And if this interests you directly, please apply using the link below. This link is a catch-all site and so the job description and salary range are generic. It's there to help us streamline collection and review of resumes. opportunities.columbia.edu/j…

Equivariance is dead😢 Or is it?😈 Genie 3 is out! Our latest protein design model yields SoTA results for binder design & motif scaffolding, greatly improving on BindCraft & Proteina-Complexa It does so using all-atom SE(3)-equivariance on a branched polymer representation👇