About
Hi! My name is Srishti and I am an ML Research Engineer.
Below are some things I’m currently the most interested in:
- Techniques for efficient training of and inference from Language Models (including data efficiency – revising how we think about synth data)
- What RL based post-training does to models? Is it helpful even when the pre-trained model hasn’t seen a lot of the data/task at hand eg. extremely low resource langs?
- Memory layouts in various tensor/array libs
- Some ML compiler stuff!
I was a Research Scholar at Cohere until July of 2025 where I worked on improving codegen models. I’ve since taken a short break and will be looking for new roles beginning Jan 2025.