Shital Shah

If universe is an optimizer, what is its loss function?

Hello there! I’m a Research Engineer at Microsoft Research with interests in deep learning and reinforcement learning.

Some of my open source works:

• I lead a team for the code infrastructure to train the Phi series of models: Phi-1, Phi-2, Phi-3, Phi-4.
• I co-created Archai, Neural Architecture Search (NAS) framework that we used to create one of the super tiny Transformer models powering the text completition feature in many Microsoft products.
• I conceived and created AirSim, a physically and visually realistic cross-platform simulator for AI research
• I conceived and created TensorWatch, a new approach for debugging training and visualization of vision models.

You can find a lot of my hobby projects on GitHub.

You can find research papers I contributed to at Google Scholar.

You can follow me on twitter for posts mainly on deep learning code and research.

About Site #

This site is made with Hugo static site generator with slighly customized Congo theme. The source code for this website is available freely. This website can also be visited at https://shitalshah.com and https://sytelus.github.io/. Pages on this website are not immutable and will be edited as neccesory.

A lot of content on this website is sourced from my twitter account. One of the goals was to move my content on Twitter/X out of its private data wall so it can be indexed and searched freely on Internet. To achieve this, I exported my data out of Twitter and wrote some code to convert it to markdown posts for this website.

Happy browsing!