AI Scientists

A framework for benchmarking AI scientist capabilities using adversarial, fast-moving real-world domains, where expert practitioners independently produce observable ground truth after an information cutoff. Instantiated in Formula 1 car design for the 2026 regulations and Magic: The Gathering deck building, with frontier models ideating from pre-cutoff information only and their outputs scored against real innovations and Pro Tour decklists.

Presented two workshop papers at ICML 2026 (a spotlight at the GenBio workshop and a paper at the AI for Science workshop) and, together with the other Encode Fellows, organised an AI for Science meet-up.

Framing oncology clinical development as an offline decision-making problem, where an agent predicts the next six-month trial portfolio of a drug programme from information available at the decision date. Built on a temporal dataset of 31.7k public records turned into 881 decision episodes across 45 historical programmes, used to train offline policies and compare them against frontier LLM agents on held-out drug, sponsor, drug-class and temporal splits.

Organised and hosted a breakfast gathering at ICML 2026 with my fellow Encode AI for Science Fellows, bringing together founders, researchers and builders working at the intersection of AI and science to discuss early-stage startups, the transition from research to commercialisation, and the infrastructure the field still needs.

Building the interface for agent-driven science: a personalised discovery feed and self-improving knowledge base for researchers and their agents.

Blog post arguing that the speed and scale benefits are obvious, but perhaps the deeper opportunity for AI scientists will be in how they communicate and make science, for the first time, genuinely social.

Chaired the ‘How I Built This: Drug Discovery’ panel at the Scientists to Builders summit at the Francis Crick Institute, with Tim Weaving (CEO, Q Matter) and Chris Thompson (CTO, Micrographia Bio). The summit brought together researchers and early-stage founders on translating AI-driven scientific discovery into real-world impact. Hosted by Conception X, Encode: AI for Science (Pillar VC) and Renaissance Philanthropy, supported by Ada Ventures and the Crick Science Entrepreneur Network, and powered by ARIA.