I am a Senior Applied Scientist, part of AWS Learned Systems Group (LSG) led by Prof. Tim Kraska, where I work at the intersection of Machine Learning and Systems. I study complex, real-world, partially-observable dynamical systems (a.k.a ‘world models’) in the context of decision making under uncertainty using offline reinforcement learning. My work primarily involves analyzing high dimensional sequential data (e.g., time series, system logs, etc) for various downstream tasks like forecasting, anomaly detection, causal discovery, inference, attribution, and decision making for two key AWS services– Amazon RDS and Amazon Redshift.

I lead the science efforts for Amazon’s Relational Database Service (RDS) team with the vision of building ‘Self-Healing’ database systems. Some of my publicly available recent work in that direction includes Amazon DevOps Guru for RDS launched at re:Invent 2021 which is an end-to-end autonomous performance monitoring service that detects performance bottlenecks, identifies root causes, and recommends fixes without any human-in-the-loop for Amazon RDS customers (Aurora and RDS engines). (Press Coverage: TechCrunch, ZDNet, User Tweets).

With Amazon Redshift, I work on the Redshift Serverless service improving its decision making to be smarter and adaptive to dynamic user workload in real-time using machine learning & reinforcement learning. The feature was previewed at re:Invent 2023.

Prior work

Before AWS, I was a graduate student at UC Berkeley advised by Dr. Anil Aswani and Dr. Ram Akella where my research focussed on Predictive Maintenance solving large scale optimization problems in scheduling for manufacturing systems. More specifically, I built model-based reinforcement learning algorithms to solve long horizon planning with delayed rewards.

Prior to that I spent time working with Dr. Ramesh Raskar and Amy Robinson Sterling at MIT Media Lab on various problems in computer vision related to real-time 2D/3D pose estimation and image segmentation. (TEDx Talk: Link).

Before that I was a software engineer at Facebook, part of the Connectivity Lab, building deep learning algorithms to solve computer vision problems supervised by Dr. Ramesh Raskar.

Outside of work

I enjoy sharing my knowledge and experiences via talks, workshops, meetups, etc. Here are some recorded ones: TEDx 2018 | Re:Work 2017 | HostingCon Cloud Summit 2017 | Coca-Cola Youth Speaker Forum 2016 | Data Natives India 2017.

I love solving real-world problems and have been an entrepreneur with successful exit in e-commerce sector in 2016. Related media coverage: Top-10 Startups in India 2015, Inc42 Magazine and VCCircle.

I enjoy playing cicket and used to play professionally at national-level representing Mumbai Cricket Association (MCA) from 2010-15.

Get in touch

Reach out via email: (my first name) @berkeley.edu