Interview with Ajinkya K Mulay

Github: @thehimalayanleo

Where are you based?  

Bay Area in the US.

What do you do (i.e. studying, working, etc.)?

I recently defended my PhD at Purdue University, where I worked on developing theoretically sound sparse recovery algorithms under differential privacy and federated learning.  Currently, I work as a research scientist at Meta.


What are your specialties (i.e. Python development, Javascript development, community organization, etc.)?

I like to work at the intersection of research and engineering. Most of my research during my PhD has focused on efficient and sparse privacy-preserving AI as well as on  computational social sciences. Although I primarily work on the development of new  learning systems, I also like building the infrastructure stacks for such systems. I prefer development in Python, C++, and PHP.

How and when did you originally come across OpenMined?

I started working in the general field of privacy-preserving AI around the end of 2019 and I came across OpenMined towards the beginning of 2020 while looking up topics for my PhD research. Four years ago, privacy was a much smaller field, and OpenMined perfectly opened up the space to meet like-minded researchers in privacy and similar domains.

What was the first thing you started working on within OpenMined?

In 2020, I was able to collaborate with several researchers (mostly, graduate and undergraduate students) on identifying the “best” Federated learning algorithm depending on the type of task at hand. During that period, FL benchmarks were few and it was hard to gauge their usability in various settings. It was a fun research project, that introduced me to several of my future collaborators while giving me a sense of the research scope in privacy. Since I was early in my PhD journey, this project helped me understand and align my interests better!

And what are you working on now?

I recently started working as a Research Scientist at Meta where I work on applied ML fields.

What would you say to someone who wants to start contributing?

I would recommend learning by doing. Working with simpler open issues on the OM GitHub can be a great starting point. Otherwise, one could pick a research paper to either reproduce or extend on their own. Once you have a decent sense of topics that you prefer to work on, consider collaborating with peers or more experienced researchers at OM.
Several conferences now host low-barrier-to-entry tracks such as Tiny Papers or ML competitions that can be a good starting point for new researchers. The community at OM is very welcoming and there are always members who can mentor you and provide you with guidance to get you unstuck along your journey!


Please recommend one interesting book, podcast or resource to the OpenMined community.

‘Peak’ by Anders Ericsson and Robert Pool