Build the toolbox for AI scientists

Own Rafflesia Homology, our serverless search engine that runs directly on object storage. You'll make tree-of-life-scale search fast, reproducible, and an order of magnitude cheaper: designing immutable index formats, squeezing every query down to the bytes it truly needs, and benchmarking hard against systems like MMseqs2 and Foldseek. Systems, storage, and search background welcome; the biology is learnable.

Own the semantic layer that lets agents move across proteins, structures, variants, assays, and papers without losing meaning or provenance. You'll design typed models, reconcile identifiers across messy public and proprietary sources, and expose it all through small, stable, versioned interfaces built for agents. For people who care about precise names, explicit types, and abstractions that don't collapse under real data.

Own AI Foundry, the typed API that puts open biological models behind one uniform surface: structure prediction, embeddings, protein design, and variant effects. You'll turn a sprawl of research repos and checkpoints into explicit model releases with content-addressed results, idempotent jobs, and provenance that survives a rerun, while keeping GPU inference fast and cheap across external providers. For engineers who like serving infrastructure, distributed systems, and making other people's research code run reliably at scale.

Turn real structural-biology research into reproducible environments that show what agents can do, where they break, and what we should build next. You'll reproduce papers from raw data, design graders and partial-credit signals for long-horizon work, and turn agent failures into usable training signal. A fit if you like reading papers, rebuilding workflows, and finding exactly where a polished result stops being trustworthy.

Build the surfaces where scientists and agents inspect data, run tools, read eval traces, and turn raw infrastructure into daily research workflows. You'll own features end to end, from data model and API contract to a fast, precise, accessible interface, making complex scientific state legible without hiding provenance or uncertainty. Strong full-stack engineers with real product judgment and a taste for interaction detail.

Embed with pharma and AI labs to turn their hardest research workflows into deployed tools, benchmarks, and evals, then fold every hard-won lesson back into the core platform. You'll scope ambiguous problems into real deployments with clear data contracts, run evaluations with partners, and generalize one-off wins into durable product instead of consulting code. For engineers who thrive in ambiguity and next to demanding technical users.