I am currently a Sr. ML Engineer at Zefr, a Machine Learning advisor and a Research Engineer at OpenMined. Previously I was a ML Engineer at InfoPlay.ai where I worked in Deep Reinforcement Learning, and prior to that I was an Applied ML scientist at various startups.
My research interests are Deep Reinforcement Learning and Federated Learning and Homomorphic Encryption.
1) I advise various startups in my free time.
2) I am a mentor to junior Machine Learning Engineers and Data Scientists.
Interest in Machine Learning
Privacy in Machine Learning
I interned and did some contract work in medically related fields for a bit and the importance of privacy in the medical field cannot be understated. I believe that accomplishing privacy in ML will be the next big step in large scale adoption as the notion that users can keep their own data private (homomorphic encryption) while still getting access to these predictive models would address the concerns that many have. My favorite paper involving homomorphic encryption and deep learning would easily be Low Latency Privacy Preserving Inference which, in my opinion, is a masterful display of the authors understanding of computer systems, cool math, and machine learning. It is my paper of the year for 2019 and the year isn’t even done yet.
I have a special place in my heart for Reinforcement Learning - for the longest time I was fascinated with the idea of creating an “artificial intelligence” and Reinforcement Learning always seemed like the closest chance we’ve had. Unfortunately, I’ve taken a step back from it for the time being - I think the constant stream of papers, difficulty of reproducing results, and slow (understandably) industrial adoption have discouraged me from continuing to pursue it as a career path.
Problem solving and physical activity! What’s not to like?
I’m interested in high performance languages and C++ seems to be the defacto language. Recently, I’ve been interested in the idea of creating a distributed a parallel computing framework a la Dask
As I get further in my career, I’ve become more interested in abstractions over my code. I tend to notice similar patterns, as well as problems in the code that I study and I think that being able to operate on a higher level and being able to guarantee correctness at a higher level is very appealing