ML Engineer

Proteus Bio; Greater London; Hybrid; GBP 75,000 - 110,000; Full time

Benefits offered by this job

Hybrid work model

Job summary

Proteus Bio in London is seeking an ML Engineer to join our Engineering team. You will help build a Biological Intelligence Platform that enables autonomous discovery of personalised medicine, reasoning over targets and proposing candidates.

The role is full-time with a hybrid work arrangement in London, starting September 2026. You should have strong ML engineering fundamentals, experience with post-training methods, and a willingness to learn biology aspects as needed.

Qualifications

• Strong ML engineering fundamentals and ability to bring systems from idea to production.
• Experience with post-training — SFT, RL, preference optimization, evals, reward models, or data pipelines.
• Biology knowledge is a plus but not required.
• Ability to work autonomously and own projects.

Responsibilities

• Build and deploy ML systems from concept to production.
• Contribute to SFT/instruction tuning, preference training, and reward-model workflows.
• Collaborate with biology and engineering teams in a hybrid London setting.

Tools

PyTorch

Transformers

Hugging Face

DeepSpeed

FSDP

Distributed training

TRL

Job description

We are hiring an ML Engineer to join Proteus in London.

We are building a Biological Intelligence Platform for the autonomous discovery of personalised medicine. Our platform reasons over disease targets, designs therapeutic and diagnostic candidates, and continuously learns from experimental results.

Location: London. Employment type: Full time.

Workplace: Hybrid.

Start: September 2026. Department: Engineering.

We are looking for someone who

• Has strong ML engineering fundamentals and can take systems from idea to production
• Has experience with post-training — SFT, RL, preference optimization, evals, reward models, or data pipelines
• Operates with a high degree of autonomy
• Ideally knows biology, although curiosity and the ability to learn quickly matter more
• Wants to leave the world better than they found it

Preferred Experience

• PyTorch and Hugging Face: Transformers, TRL, Accelerate
• SFT / instruction tuning
• LoRA, QLoRA and PEFT
• DPO / g-DPO for preference training
• GRPO / PPO for online reinforcement learning
• Reward models, rule-based verifiers and outcome-based rewards
• Tool-use trajectory collection and credit assignment
• Distributed GPU training: DeepSpeed, FSDP, mixed precision
• Model evaluation, checkpointing and experiment tracking

Biology experience is valuable, but not required. We care more about what you have built, how you think, and whether you take ownership than where you went to school.

We are backed by Founders, Inc.