Talks & Events
Prior to 2024, most talks I gave were internal to Microsoft and Amazon. Here is a set of recent public facing talks and invitations.
Talks
2025
- Compositional Rules for Singular Learning Theory - 5th September, Department of Mathematics, Imperial College London, UK.
- GCHQ CyberFirst Programme, The current state of AI Security, August 1st, Microsoft London, UK.
- Applied Topology Conference, Persistent Homology for LLM Interpretability, July 14-18th, Poznań, Poland.
- Holes in Latent Space, Topological Signatures of Adversarial Influence , London Oxford Topological Data Analysis Seminar Day, May 22nd, London, UK.
- 2025 Enhancing Capabilities for Detecting Adverse Cyber Activities, United Nations Cyber Norms Workshop, May 19–20th, Berlin, Germany.
- IEEE Secure and Trustworthy Machine Learning Conference. Two contributed talks (main track and competition track). April 9-11th, Copenhagen, Denmark.
- Interpreting LLM System Behavior with Basic Math and Statistics, GirlsWhoML workshop, Entrepreneur First, February, London, UK.
- Zero Day Quest: Microsoft’s Bug Bounty Program and AI Research, January (online).
2024
- "The Long View on AI: What’s Next and What Matters" Panel, AI Fundamentals for Civil Servants, December, Imperial College London, UK.
2023
- Breach Path Busters: Decoding Cyber Threats using Graphs and AI, Microsoft BlueHat, Washington, United States.
- United Nations Internet Governance Forum, Kyoto, Japan.
- 2023 NATO European Cyber Conflict Research Initiative, Brussels, Belgium.
Research Visits
2025
- Microsoft Zero Day Quest On-Site Hacking Event for AI Security vulnerabilities, Washington, United States.
- Erlangen AI Hub Conference, Queen Mary University London.
- Erlangen AI Hub Year 2 Kick Off, University of Oxford.
Poster Presentations
2025
- Holes in Latent Space Poster presentation acceptance at Applied Topology Realization Network conference, Chicago, USA, August.
- Erlangen AI Mathematical Foundations of AI Conference 9-11 June.
- London School of Geometry and Machine Learning Summer School.
- IEEE Secure and Trustworthy Machine Learning Conference. Two contributed talks and posters.
Summer Schools
- London School of Geometry and Machine Learning Summer School Acceptance to work on Cycle Matching for LLM Interpretability using our TaskTracker dataset. 7-11 July.