I'm Salma. I work on reinforcement learning and large-scale optimization at Google DeepMind. I am building noether. Outside of core infra, I've started collaborating with Sara Seager's group at MIT on exoplanet habitability modeling.
Before DeepMind, I built multi-agent RL pipelines for incident triage at Microsoft Research. Before that, I worked on privacy and federated learning at OpenMined, and earlier, interned at Microsoft's Experiences + Devices org.
I like designing and reasoning about systems. Almost anything—software, biology, institutions—can be modeled as one.
The best way to get into my mind is to read .
In my free time I read a lot, write poetry and listen to music almost all of my waking hours. I pick up new skills and hobbies compulsively. Over the years, I've learned how to play badminton, the keyboard, paint, sketch, origami, horse riding, swimming, speak 4 languages (Russian, German, Korean, Japanese), and more recently Muay Thai (amateur; picking it up!). It's surprising how fast you get better at something with sustained effort; and how much learning one thing compounds into another.
Strong opinions, weakly held
- 1. Writing is the cleanest way to catch yourself thinking.
- 2. Half-formed thoughts often say more than polished arguments.
- 3. Impatience, audacity, and persistence are underrated; together, they're a better heuristic than most productivity hacks.
Contact
Write. Create. Recurse.