Hello there, I am Vinay Sisodia. I have deep interest in areas of math & machine learning. I am currently working as a Machine Learning Tech Lead at PicCollage.

What I do

• Bringing AI to a tech startup: building ML technologies, systems and infrastructure from the ground up
• Building and growing a machine learning team
• ML Research, esp vision and language models: I am particularly interested in multi-modality, self-supervised learning and graphs
• Building new products

You can learn more about my technical projects here.

Things I have built

• Linear Algebra for Programmers: An e-book of visual essays to learn linear algebra from scratch.
• TinyVolt: A collection of interactive visual essays related to topics in math. Vastly inspired from The Pudding and, to a certain extent, Instill.
• Normalizing flows - theory and implementation: A YouTube playlist introducting the ideas behind normalizing flows and how to implement them in Python.
• Distillation of CLIP model (Medium article): Work done at PicCollage. The article covers only the initial attempt at distillation of the CLIP model. The work was then extended to extraction of attention values for image patches and re-training the distilled model with the attention values.
• Category Theory Meetup: A (now defunct) meetup group for discussing topics related to Category Theory. At some point I was really excited about this topic but soon realized that unless I familiarized myself with some advanced topics in abstract algebra, it would sound and feel like abstract nonsense. Incomplete notes on this topic can be found here.

Some background

I got into machine learning when companies were starting to build dedicated AI/ML teams. I had the privilege of being the first ML engineer in two different organizations, which allowed me to build ML systems from the ground up. Later on I took on the responsibility of growing and leading ML teams. The past few years have been quite gratifying professionally and intellectually, and I am very proud of some of the things I built during this time.

I love taking notes

I have been trying to organize them and put more of my notes in public domain (using HackMD). Check out the list of all the notes available there under the Notes with code section.

A small fraction of my notes can be found in these Github repos - Understanding neural networks with PyTorch and Math and ML notes.