Adaptyv is building an automated lab thats let AI agents run biology experiments.
We're entering the era of agentic science where AI models can now design novel proteins, propose hypotheses, and iterate on experimental results. But they can't run the experiments themselves - that's still a manual, months-long process. We're building the infrastructure that gives AI agents access to the physical world.
We are one of the fastest growing biotech companies, trusted by leading biopharmas, frontier AI labs, and the techbio companies pushing the field forward. This is a rare chance to help advance some of the most important work happening in biotech today.
Our automated lab is powered by a deep software + hardware stack: lab instruments worth millions of USD reverse-engineered into API-controllable hardware, dozens of devices orchestrated through complex workflows, full observability on everything that happens in the lab, processing pipelines for messy physical-world data, and AI systems that troubleshoot production results and accelerate assay development.
We’re growing rapidly and are hiring for talented people to scale and support the massive demand for AI-driven wet lab experimentation.
You'll build out the data science layer of Adaptyv's foundry — the work that turns tens of thousands of raw, messy experimental readouts into clean, trustworthy, structured data that our customers, our models, and our own scientists can rely on. Binding (BLI/SPR), developability, biophysical, and functional assays all produce data at scale; your job is to make that data correct, comparable, and useful.
This sits at the intersection of three things: data quality (is this number real, or an artifact?), bioinformatics (linking experimental results back to sequence, structure, and protein design), and dataset building (turning foundry output into the kind of high-quality, benchmarkable data that frontier AI labs actually want). You'll work shoulder-to-shoulder with the lab scientists who run the assays, the software team who own the pipelines, and the customers who train models on what we produce. This is a hands-on build role, not a management one.
Own the scientific logic of data quality across the foundry: define what "good data" looks like for each assay type — expected signal ranges, control thresholds, failure modes, edge cases — and turn it into automated checks.
Build anomaly detection and QC models that catch bad data the eye would miss: assay drift, instrument variability, plate effects, false passes and false fails — and distinguish real signal from noise statistically.
Work with the software and ML teams to specify, review, and improve the automated data pipelines that process instrument outputs, feeding back precise requirements for what to flag, auto-reject, or route to human review.
Connect experimental results back to the protein side — sequence, structure, and design — so wet-lab data and computational models reinforce each other.
Turn foundry output into structured, documented, benchmark-grade datasets that are a genuine asset for our customers and for training and evaluating protein-design models.
Apply real statistical rigor to multi-condition data at scale — thousands of samples across hundreds of simultaneous experiments — and make the results interpretable and comparable across runs.
Strong data science / bioinformatics background — you're fluent in Python (pandas, numpy, the scientific stack) and comfortable owning messy, real-world experimental data end to end.
Genuine biology grounding — you understand proteins, assays, and sequence/structure/function well enough to know what the data means, not just how to process it. You don't need to be a bench scientist, but you can't be biology-blind.
Statistical maturity — process control, anomaly detection, handling variability and batch effects; you can tell drift from noise and defend the call.
Prolific builder with the receipts to prove it. You've shipped a lot — pipelines, tools, models, datasets — and can point to concrete things you built end to end and put into real use, not prototypes that died in a notebook. You move fast, systematize what works, and have no patience for babysitting a fixed dashboard.
AI-native builder. It's 2026 — you build with coding agents like Claude Code as a default, and you have sharp judgment about what they produce.
Interdisciplinary by instinct. You're energized working across the lab bench, software, and ML, and you treat automation and data infrastructure as part of your job.
Bonus: experience with protein/sequence-structure data (bioinformatics tooling, structural data), ML on experimental data, or building datasets for model training and benchmarking.
Application deadline
We are reviewing applicants on a rolling basis.
