About
I am an AI safety research engineer with a background in ML research and engineering. I have worked on evaluations of frontier LLM agents and am broadly interested in technical approaches to AI safety, motivated to help ensure that increasingly capable AI systems are developed safely.
Recent Activity
Apr 2026 – Present
Research Engineer at AI Safety South Africa, continuing and extending my LLM-agent evaluation research through the development of low-cost, predictive proxy evaluations.
Apr 2026
Teaching assistant for the ML4Good AI safety bootcamp.
Oct 2025 – Mar 2026
Research Engineer (contract) with the Science of Evaluations team at the UK AI Security Institute (funded by the UK AISI Challenge Fund), building evaluations that measure the adaptability of LLM agents as a predictor of real-world performance.
Mar 2025 – Sep 2025
Worked on the “Precursors to Dangerous Capabilities in AI Agents” project with Leo Hyams and Sam Brown, which led to a poster, “Precursors, Proxies, and Predictive Models for Long-Horizon Tasks”, at the NeurIPS 2025 LLM Evaluation Workshop.
Jan – Apr 2025
Worked with AI Safety Engineering Taskforce as part of the UK AI Security Institute Bounty program, where all four of my agent evaluations were accepted into the UK AISI internal evaluation suite.
Background and History
I studied Mathematics and Computer Science at Stellenbosch University in South Africa, then did my Master’s in Computer Science there, where I worked on fine-tuning encoder-based language models for classification tasks.
Other Interests
Apart from working on AI, I enjoy:
- Reading
- Running, Cycling, and Hiking
- Traveling
- Paragliding