hi! i’m dron. while i’m writing this, i’m a third-year computer science student at the university of cambridge, and i’m looking for my next adventure post-graduation. i wonder what’ll happen once i stop writing this!
things i’ve done
Most of my experience is in ML research, in and around AI safety and robustness. I’ve worked on:
- Generative world modeling for inferring 3-D scenes from 2-D images at the Center for Human-Compatible AI with Shreyas Kapur (Berkeley)
- Scalably testing worst-case adversarial robustness in computer vision with Dan Hendrycks (Center for AI Safety) and Max Kaufmann (UK AI Safety Institute)
- Investigating universality of Sparse Autoencoder bases with Davis Brown (UPenn)
In the past, I also worked with distributed systems at Zilliqa and did some weird stuff at X. I have a somewhat amateur fascination with generative models for music, and I used to be mediocre at math olympiads, back when I got the median score at the Canadian Math Olympiad. Here’s my resume.
things i’m doing
I’m currently a Research Fellow at Goodfire, where I’m training interpreter models like SAEs at scale.
I’m also working on making an actually good GPU cluster management system to free ML researchers from the shackles of Slurm. Or, if you prefer, Ergonomic GPU Dispatch and Allocation for Machine Learning Research, which is the working title of my undergrad thesis. I’m also learning category theory, auditioning for big bands, playing tennis and volleyball, and thinking about what to do next year; I’m trying to enjoy my last year of college.
things i want to do
In the past, I’ve insisted on defining specific problems that I wanted to work on: I wanted my interests to be concrete and well-defined. Doing this for the last few years has been fun! It’s the main way I’ve turned “things i want to do” into “things i’ve done.” But now, having wrapped up most of my past work, I don’t have an obvious correct next step. Instead of optimizing for legibility by choosing a concrete direction roughly at random, I’m admitting my ignorance: I want to explore, learn, and train.
That said: I’m most interested in research engineering roles, where I can spend a lot of time close to The Iron running and iterating on ML experiments. I’m actively looking for full-time or intern roles starting after June 2025, which is when I plan to graduate. I’m not beholden to any particular field or research direction, though I have the most direct experience working with vision and perception. I think I’m capable of getting up to speed on any direction quickly:)
If I don’t care about direction, what do I care about?
I want to build my research skills bottom-up, working through and understanding details, turning ideas into experiments that can be iterated on. I’ve learned the most from the people I’ve worked with in the past, so what I care about most is the people I’d work with in the future. I think I would most enjoy:
- Working at larger scales (both compute and people), contributing as part of a larger team, training engineering skills and ML debugging intuition
- Working at the smallest scale, spending extensive time with a single mentor, learning research heuristics by osmosis
Basically: I want to focus on training my low-level execution skills and let the environment and my subconscious generate some higher-level insights in the background.
If any of this seems interesting to you, please reach out! Admin details:
- Email: dronh.to at the google mail website
- Current location: half Cambridge (UK), half Toronto
- Future location preference: anywhere in US/UK/Canada, modest preference for London or US east coast