Founding Research Scientist / Engineer: User Modelling, LLM Post-Training & Evaluation

Stealth Startup; Greater London; On-site; GBP 90,000 - 150,000; Full time

Benefits offered by this job

Equity 1.2%–3%

Job summary

Stealth Startup in London is building an early-stage AI company focused on adaptive intelligence and reinforcement learning. You’ll help shape the research direction, core technology and engineering foundations with substantial autonomy to identify important problems and lead initiatives, with potential for leadership roles as the company grows.

Depending on your interests, you’ll contribute to adaptive intelligence, RL and post-training, continual learning, evaluation and generalisation, and

Qualifications

• Strong ML foundations with model architectures, training methods and evaluation.
• Hands-on programming in Python and ML frameworks.
• Experience conducting ML research or building sophisticated ML systems.
• Ability to work independently with limited direction.
• Scientific rigour and willingness to challenge existing approaches.
• Genuine interest in building a company, not just a research project.
• 2+ years of experience at a frontier AI lab or high-impact startup.

Responsibilities

• Shape the research direction, core technology and engineering foundations.
• Lead research initiatives and turn promising results into real-world systems.
• Collaborate with founders to shape technical culture and strategy.
• Potential to take on research or engineering leadership as the company grows.

Skills

ML foundations

Python

PyTorch

Research leadership

Education

PhD or MSc in AI/ML

Tools

TensorFlow

JAX

Job description

We’re an early-stage AI company, currently in stealth, tackling one of artificial intelligence’s fundamental challenges: making AI systems genuinely adaptive. Foundation models have made extraordinary progress, yet they still struggle to adapt to new tasks, environments and changing requirements. We believe the next major advances in AI will come not only from making models more capable, but from enabling them to learn, adapt and improve through experience.

Our founding team brings together experience from Google DeepMind and Oxford, with deep expertise in machine learning, reinforcement learning, foundation models, adaptation and personalisation. At DeepMind, our team members worked on Gemini and led research on user adaptation, evaluation and model behaviour.

Our scientific advisors are internationally recognised researchers in artificial intelligence: Michal Valko, co‑founder of DeepMind’s Paris office and Professor Philip Torr, University of Oxford. We also collaborate with the University of Cambridge on computational models of cognition and human states. We’ve also secured early backing from venture capital investors and angels.

The Opportunity

You’ll join us at a formative stage and help shape the research direction, core technology and engineering foundations of a new AI company. You’ll have substantial autonomy to identify important problems, develop novel methods, lead research initiatives and turn promising results into real‑world systems. As the company grows, you’ll help shape its technical culture and strategy, and can take on research or engineering leadership.

Depending on your expertise and interests, you’ll contribute to areas including:

• Adaptive intelligence: new approaches that let AI systems adapt to changing tasks, environments and requirements.
• Reinforcement learning and post‑training: algorithms that improve the capabilities, adaptability and behaviour of foundation models.
• Continual learning: methods that let AI systems learn from interactions, feedback and outcomes.
• Evaluation and generalisation: rigorous methods for measuring adaptation, learning, robustness and generalisation.
• Research to production: turning fundamental research into scalable, real‑world AI systems.

Who We’re Looking For

We’re particularly interested in people who combine strong machine learning expertise with an entrepreneurial mindset, and who thrive on difficult, open‑ended problems.

You should have:

• Excellent foundations in modern machine learning, with a strong understanding of model architectures, training methods and evaluation;
• Strong hands‑on programming skills, particularly in Python and PyTorch, JAX or equivalent frameworks;
• Experience conducting ML research, building sophisticated ML systems, or both;
• The ability to work independently and solve difficult problems with limited direction;
• Scientific rigour and a willingness to challenge existing approaches;
• A genuine interest in building a company, not just contributing to an isolated research project;
• 2+ years of experience at a frontier AI lab or high‑impact startup.

Particularly relevant experience:

• LLM post‑training, reinforcement learning, reward modelling or preference optimisation;
• Personalisation, recommendation systems, user modelling or long‑term memory;
• AI agents, adaptive systems or human‑AI interaction;
• Scalable ML infrastructure, distributed training or inference systems;
• Publications at top conferences (ICLR, NeurIPS, ACL, EMNLP, etc.);
• PhD/MSc in AI/ML.

Location & Compensation

• Location: London, King’s Cross, UK
• Type: Full‑time
• Equity: 1.2–3%
• Both depending on experience and responsibilities.

We’re looking for people excited to help build a new AI company and make a lasting impact. If that sounds like you, we’d love to hear from you.