Salma Shaik

I'm the founder and CEO of Noether Labs. We build runtime safety for AI that controls physical machines, starting with precision optics. Separately, I collaborate with Sara Seager's group at MIT on exoplanet habitability modeling (prev.).

Before Noether, I spent a year at Microsoft Research (Systems Innovation) teaching groups of agents to triage incidents in the cloud. We got one of the first RL-informed reliability systems running at production scale, and much of what I'm building now grew out of the questions it left behind. Before that, I worked on privacy and federated learning with OpenMined, an open-source research collective, and interned at Microsoft.

Current research hyperfixation: can a differentiable optical model run fast enough (~1 ms) to sit inside an adaptive optics loop, and where should a learned controller take over from the physics? More broadly, I think about how learned systems behave once they're wired into real machines, and how to verify they actually did the thing.

I studied computer science and machine learning, and wandered off often, into computational biology (CMU), robotic perception (Artenal, a startup building harvester robots, back in the day), logic, physics, and the philosophy of consciousness.

I tend to pick up new skills and hobbies frequently. Over the years, I've learned badminton, piano, painting, sketching, origami, horseback riding, swimming, belly dancing, ikebana (and chahua) flower arrangement, speedcubing (ongoing), and several languages (Russian, German, Korean, and Japanese). More recently, I've been learning Muay Thai. It's surprising how quickly you get better at something with sustained effort, and how much learning one thing compounds into another.

You can find me on X and GitHub.

You can email me at salma [at] noether [dot] one.

I try and respond to most email but it really helps if you follow this advice.

The best way to reach out however would be getting an introduction via someone I already know.

I am based in Bangalore.