Product ML Engineer: Post-Training for Open Models

PhysicsX Ltd.; City Of London; On-site; GBP 90,000 - 130,000; Full time

Job summary

PhysicsX Ltd. in London is seeking a hands-on product ML engineer embedded with our Research team to turn post-training methods into production-ready capabilities.

You will drive improvements for open models and shape a repeatable platform that lets customers train on their own workflows while protecting data and IP. You will own the capability roadmap, apply post-training methods (SFT, RLHF, distillation), build synthetic data pipelines, and ensure measurable gains on our benchmarks.

Qualifications

• Hands-on ML engineer with experience in post-training methods.
• Ability to define success and deliver measurable improvements.
• Comfort working in cross-functional product teams.
• Experience turning research into production-ready features.

Responsibilities

• Own the path from a base model and customer workflow to a deployable improvement.
• Build post-training harnesses covering data preparation, training, evaluation and artifacts.
• Improve open models against benchmarks using evidence to choose training vs prompts.
• Develop methods across SFT, RLHF/RLAIF, distillation, synthetic data and verifiers.
• Create reproducible data pipelines with quality controls and provenance.
• Ensure post-training capabilities remain on customer platforms across deployments.

Skills

Python

Experimentation

Product ownership

Cross-functional collaboration

Tools

PyTorch

ML experiment tracking

Job description

PhysicsX Ltd. in London is seeking a hands-on product ML engineer embedded with our Research team to turn post-training methods into production-ready capabilities.

You will drive improvements for open models and shape a repeatable platform that lets customers train on their own workflows while protecting data and IP. You will own the capability roadmap, apply post-training methods (SFT, RLHF, distillation), build synthetic data pipelines, and ensure measurable gains on our benchmarks.