Hi there.
I’m an ML engineer and technical leader with 13 years experience in the UK government and early-stage ventures. I train neural networks that learn vehicle physics and driving, to provide feedback to drivers globally. I’m now focused on cheap, scalable monitoring of LLMs, including distilled NLAs combined with classifiers to detect evaluation awareness and deception from model internals in real time.
We're living in a critical time in history. I feel both deep excitement and urgency to make tools to better understand and monitor the minds we're building.
If you want to help me with this you can:
- Introduce me to new peers and collaborators
- Point me to conferences
- Critique my takes
- Connect me to a mentor
- Connect me to sponsors or potential users of the work I'm excited about
- Meet me in London for coffee
Contact me at adam_bricknell@live.com or LinkedIn. Or read my CV.
Research
Live doc of things I’m excited to explore in interpretability