About the roleAs a Member of Technical Staff, ML Product Engineer, this role owns the layer between the models and the people using them: the APIs, batch and compute systems, and services that make inference fast, reliable, and cheap at genome scale. You would build for our scientists and for the partners and developers who build on Omnii, our next-generation genome language model.
You care about throughput, latency, cost, and uptime, and you have real opinions about what a good developer experience feels like. The platform you build is how Omnii reaches the people using it to advance human health and biosecurity.
This role sits close to our product and data-partnership functions and our modeling team. You translate between what the science needs, what partners can consume, and what the infrastructure can deliver.
What you'll do- Build the inference APIs behind Omnii: real-time serving plus large batch scoring for genome- and variant-scale workloads, where a single job can be millions of sequences, latency-tolerant and cost-sensitive.
- Design the compute and orchestration layer: job queues, autoscaling GPU inference, retries, fault tolerance, and the observability to run all of it against real SLAs.
- Build the developer-facing product: the API surface, SDKs, and documentation that let an outside team depend on Omnii without hand-holding.
- Drive down cost and latency per inference, and make performance predictable enough to price and guarantee.
- Package the platform to run inside partner environments that cannot let data leave, including on-prem and air-gapped installs.
- Work directly with our scientists and partners to shape the API around real inference workloads.
What we're looking for- You have built and operated backend or distributed systems at production scale, and you owned their reliability.
- You have shipped a model as a service that other people depended on. You think in inference, serving, and throughput.
- You have built asynchronous or batch job systems at scale. Genome-scale workloads should feel like familiar ground.
- You write production code in Python and at least one typed language, and you are comfortable with containers and infrastructure as code.
- You have product judgment. You can decide what to expose, what to hide, and how an API should feel to the person calling it.
- You operate well with ambiguity and want the ownership of an early technical role.
Nice to have- GPU inference internals: vLLM, TensorRT-LLM, or Triton, with hands-on performance tuning.
- Experience deploying software into locked-down environments such as enterprise VPC, on-prem, or air-gapped.
- Familiarity with genomics or computational biology, or a real appetite to learn it fast.
- Experience standing up an external API product from the first endpoint forward.
