Members of Technical Staff (AI)

Location: London, United Kingdom; on-site, working in person every day

Company

hirify.global is a well-funded, fast-growing frontier AI research lab developing recursively self-improving AI for scientific discovery and AI-enabled science.

What you will do

• Design, implement, and tune supervised fine-tuning and reinforcement learning algorithms for foundation models with open-ended scientific research capabilities.
• Build autocurricula, judges, harnesses, and evaluation pipelines that convert open-ended research tasks into reliable reward signals.
• Run large-scale experiments on state-of-the-art hardware and analyse results to define subsequent research hypotheses.
• Develop recursive self-improvement loops in which AI agents contribute to their own post-training research.
• Collaborate with Infrastructure and AI for Science teams to optimise hardware and deliver performance in scientific domains.

Requirements

• 3+ years of deep learning research experience and 5+ years of software engineering experience.
• Experience post-training large language, vision, video, or multimodal models.
• Deep familiarity with Python and at least one deep learning framework, such as PyTorch or JAX.
• Demonstrated deep learning research achievements through papers, model releases, open-source contributions, or comparable work.
• Experience with modern coding agents and strong opinions on effective workflows.
• On-site availability in London is required.

Nice to have

• PhD in mathematics, computer science, or a hard science discipline.
• Hands-on experience training large language models with reinforcement learning at scale, including GRPO, PPO, DPO, or distillation.
• Familiarity with distributed and long-context training infrastructure.
• Background in autocurricula, open-endedness, meta-learning, or recursive self-improvement.
• Experience post-training frontier models at an industry lab.

Culture & Benefits

• Opportunity to shape the core research of a frontier AI lab from its beginning.
• Work on recursive self-improvement and AI scientists that improve the research pipeline that trains them.
• Directly dogfood post-trained agents to accelerate further research.
• Small, high-trust team with minimal bureaucracy and a strongly technical culture.
• Emphasis on experimental organisational design, agent-centric workflows, and diversity of thought.