Lucy Farnik

I’m an ML PhD student researching robustness and continual learning in LLM agents. I’m fortunate to be advised by Yuki Asano and Oana-Maria Camburu, alongside my official supervisors Conor Houghton and Mengyue Yang.

I started coding at age 7 and by age 18 I was a senior dev at a startup. After a few years working there, I decided to switch to ML research. During undergrad I collaborated with researchers from Oxford, Berkeley, and FAR AI. I then did ARENA, MATS (mentored by Neel Nanda from Google DeepMind), and lots of other research placements. I’ve also received over $100k in research grants and I cofounded a small research org working on AI security and emerging tech policy. My Erdős number is 4.

I’ve published at ICLR/ICML in areas ranging from RL theory to LLM interpretability. Most recently, I’m excited about ways that LLM agents can effectively learn from their open-ended interactions with their environment, and how we can ensure that the things they’re learning are aligned with what we want.

I’m always excited to meet new people with similar research interests! My email is {firstname}{lastname} at gmail. You’re also welcome to send me anonymous advice/feedback.