Research Scientist (Post-Training)

About the Company

We’re partnered with a very heavily funded early-stage AI research company, set up to apply frontier AI to engineering and industry: the design, building and making of real things.

It is small and research-led. Hiring runs on referral and research reputation currently, and the calibre of the team is incredible and the researchers are some of the strongest in the field.

The company keeps a low profile, but the ambition is huge. It wants to build AI that can engineer, not just chat about engineering.

The Role

As a Research Scientist (Post-Training) you will shape how powerful generative models become useful, reliable and genuinely capable. Pre-training gives a model its raw knowledge. Post-training decides what it can do with that knowledge, and you will be one of the people deciding.

You will work on the methods that turn a strong base model into a system that can reason, act and improve, drawing on techniques such as supervised fine-tuning (SFT) and reinforcement learning (RL) alongside generative modelling. The work is hands-on and research-first, in a team where ideas get tested quickly and good ones carry real weight.

Key Responsibilities

• Design and run post-training research that improves the capability, reasoning and reliability of large generative models
• Develop and refine SFT and RL approaches, and test them rigorously against problems that matter in the physical world
• Build the evaluations and experimental setups that tell you whether a model is actually getting better, not just looking better
• Work closely with fellow researchers and engineers to carry results from early idea to something that works at scale
• Keep up with the research frontier, and bring your own original ideas and direction to the team
• Take real ownership of your research agenda, with the freedom to follow promising threads

Essential Requirements

• A strong research record in machine learning, shown through publications, open-source work or other credible, visible contributions
• Hands-on experience with post-training methods for large models, especially SFT and RL
• A solid grounding in generative modelling
• The ability to design careful experiments, interpret messy results honestly and iterate fast
• Comfort working independently on open-ended problems, and collaborating well with other strong researchers

Why Join

• A mission with weight: you would help build AI aimed at engineering and industry, where models have to be right about the real world
• Serious backing: multi-billion funding means the ambition has the resources to match it
• Exceptional colleagues: a team assembled through research reputation and trusted referrals
• Real ownership: a flat, title-light structure where your ideas shape the direction of the research
• Early-stage influence: the chance to help set the methods, standards and culture of a lab that is still taking shape

If your research already speaks for you and this sounds like the kind of problem you want to spend the next few years on, get in touch. We can have a relaxed, confidential chat about the role, the team and whether it fits what you want next.