Senior Machine Learning Engineer, GPU Optimization

On-site · London (United Kingdom) · Senior · Full-time

Want offers like this one?

Create your free PyHires profile: you will see how well you match each offer and get a weekly email with the ones that fit you.

Create my profile

Overview

Work mode
On-site
Location
London (United Kingdom)
Engagement
Full-time
Seniority
Senior
Languages
English
Open to candidates from
All of Europe (EU and non-EU)
Company
Boltz
Posted
Nov 19, 2026

About the role

Boltz, an MIT spin-out and public benefit company, builds open molecular modelling models (Boltz-1, Boltz-2, BoltzGen) used by 100,000+ scientists, and Boltz Lab, the platform where partners design and rank candidate molecules. As ML Engineer in GPU Optimization you will own how fast and reliably the structure models train, embedded with a research team, with a near-term goal of a 1.5x speedup on the structure model. What you'll do - Speed up large-scale training by removing bottlenecks in kernels, memory, communication, data loading and parallelism - Write and tune custom GPU kernels (Triton or CUDA) where they pay off - Make long training runs complete without interruption and fix root causes of degraded runs - Build reusable profiling and optimisation tooling, including agent-assisted, profile-guided optimisation What you bring - Experience optimising deep learning training on multi-node GPU clusters (FSDP, DeepSpeed, Megatron) - Deep GPU performance expertise: profiling (PyTorch Profiler, Nsight), memory and communication tuning, Triton or CUDA kernels - Strong PyTorch and Python, with ML intuition about how engineering changes affect training - A track record of measurable speedups or reliability gains on large training workloads What they offer - One of the most talent-dense teams in the field - A highly competitive salary with meaningful equity in a public benefit company - 25 days of annual leave plus bank holidays

Requirements

Based on:Python

Must-have

  • PyTorchEssential

About Boltz

Boltz is a public benefit company building AI-powered molecular modelling tools to make biology programmable and accelerate drug discovery. It began as an MIT spin-out: its open models Boltz-1, Boltz-2 and BoltzGen are used by more than 100,000 scientists and in programmes at every Top 20 pharma, and its commercial platform Boltz Lab runs its latest models and design agents for partners.

Stack

  • Python
  • PyTorch

Company

Boltz logo
Boltz
United Kingdom
View company profile