Research Engineer, Machine Learning
Boltz · London
Bio & Science AIremote
At a glance
Highlights
- Scale foundation model training powering next-gen AI molecular modeling tools.
- Own high-impact systems work end to end alongside ML research teams.
- Make systems work reusable across teams to inform experiment design.
What they offer
- Meaningful equity in a public benefit company
- 25 days plus bank holidays a year
- Annual team offsite trip
- Private healthcare (BUPA), covering pre-existing conditions
- Pension program
- Paid parental leave
- Wellhub gym membership
- Free lunch every day
- Travel expenses
Ideal candidate
A systems-minded ML engineer who loves working close to research, independently finding high-leverage scaling problems and owning them with high engineering craft.
Requirements
- Hands-on experience training large deep learning models on multi-node GPU clusters
- End-to-end knowledge of ML training stack from systems perspective, with depth in at least one layer
- Strong Python and PyTorch with high bar for engineering quality
- Experience working closely with ML researchers on model and training code
- Track record of identifying and driving high-impact technical work independently
- Knowledge of parallelization strategies including FSDP/ZeRO, tensor and sequence parallelism
- Low-level GPU performance work such as custom CUDA or Triton kernels
- Experience building tools for shared GPU cluster use
- Exposure to structure prediction or molecular modeling or Boltz models
- hiring process
- Initial screen with Talent team, diagnosis interview working through a degraded training run, onsite in London or virtual onsite, followed by offer and onboarding.
Lab profile
About Boltz
Building AI-powered molecular modeling tools to make biology programmable and accelerate drug discovery.
What they build
- Boltz-1 open-source structure prediction model
- AI-powered molecular modeling tools
- Boltz Lab
Known for
- AlphaFold 3-level accuracy for biomolecular complexes
- Fully open-source and commercially available models
- Broadly accessible frontier capabilities
Posted · Last seen