I'm Freddie. I work on the research team at Databricks, continually improving data agents.
Previously co-founder & CTO at Quotient AI (acquired by Databricks), at GitHub working on Copilot — evals & infra for tab completion and Chat — and open research with Cohere Labs on the Aya Project.
research
| Venue | Paper | Contributions |
|---|---|---|
| Nature 2026 | Humanity's Last Exam | Contributed a difficult math & statistics question about theoretical max damage in Old School Runescape |
| ACL 2024 (Best Paper Award) | Aya Model: An Instruction Finetuned Open-Access Multilingual Language Model | Data Engineering to create the Aya Human Annotated set of the Aya eval suite: dataset |
| ACL 2024 (1st Author) | Aya Dataset: An Open-Access Collection for Multilingual Instruction Tuning | Designed, built, and scaled out the backend for an open data annotation platform for over 3000 contributors and 300,000+ annotations for training multilingual instruction-tuned LLMs on mixed-resource languages: repo. Co-created the Aya Score to encourage participants to incorporate more edits during annotation, increasing average annotation length by >50% |
| CSCW 2017 | Crowd Guilds: Worker-led Reputation and Feedback on Crowdsourcing Platforms | Helped design a reputation system for workers and requesters to gain experience, receive higher wages and higher quality work |