Lead Applied AI Researcher

• Energy is a rare place to apply AI across the full stack: from helping customers and building software to operating physical infrastructure and trading power. Our ambition is to make energy cheaper and more abundant with AI, while building the energy and compute infrastructure that the next generation of AI will need
• Design experiments to find the best model and approach for each workload, balancing accuracy, latency and cost
• Own our evaluation framework: build high-quality datasets and golden answers, combine human and LLM judgments and measure how much confidence to place in each evaluator
• Explore post-training methods, including supervised fine-tuning and reinforcement learning
• Build agent harnesses that make long-running tasks reliable: tool use, deterministic checks, state tracking, retries, resource allocation and appropriate human approvals
• Set evaluation and escalation thresholds according to the consequences of failure
• Work with teams across Fuse to turn research results into production systems and bring new methods and hard-won lessons from the frontier of applied AI into the company

Preferred experience

• Hands-on experience with several of model selection and routing, fine-tuning, reinforcement learning, agent systems or computer use
• Good judgment about cost, latency, reliability and risk, especially when deciding where deterministic checks or human review are needed
• Strong software engineering skills, including Python and the ability to build reliable experimental and production tooling
• Ability to move between research and implementation: formulate a hypothesis, run a rigorous test, inspect failures and ship what works
• Deep understanding of evaluations, dataset quality and failure analysis. You know how to distinguish a genuine improvement from an unreliable benchmark result
• Strong experience applying modern AI models to real-world tasks with a track record of improving measured outcomes through experimentation
• Experience at a frontier AI lab or a team operating at a similar level of experimentation would be especially valuable