Intent-Driven Multi-Engine Observability Dataflows For Heterogeneous Geo-Distributed Clouds
- 2024
- CLOUD 2024
I currently lead a team of researchers working on AI platform topics and I lead GenAI model customization globally for IBM Research. My team works on various aspects ranging from model customization techniques and recipes, platform and middleware tools, building the software stack for IBM's AIU Spyre accelerator, distributed inference and compliance. We have developed an open-source fine-tuning stack based on PyTorch and HuggingFace libraries, along with optimized kernels (40-100% throughput gains) and ease-of-use features, that is incorporated as part of Red Hat Openshift AI and IBM watsonx.ai products, and used by 20+ Research teams developing customized GenAI models for various IBM products.
In my previous stint, I led work on telecom, 5G and edge computing, with a focus on aspects of observability, intelligent operations and management of network functions and enterprise workloads running as containers on Kubernetes. I made significant contributions to IBM Hybrid Cloud Mesh, helping shape the product from inception. Prior to this, I lead a team of researchers in the Blockchain and Smart Contracts Group at IBM Research - India. Our efforts were focussed on developing first-of-a-kind enterprise-grade blockchain solution networks (TradeLens) reimagining industry use cases in a blockchain world, contributing core platform capabilities to Hyperledger Fabric, blockchain interoperability and other value added services. Prior to that I was a part of the Cloud Computing Research Group.
I co-taught an online blockchain course together with Prof. Sandip Chakraborty on the NPTEL platform for 3 semesters in 2018-19. The course had over 50000 registrants over the three editions and was covered in the press.
I obtained my PhD from the University of Illinois at Urbana-Champaign in September 2010, advised by Prof. Tarek Abdelzaher. My PhD thesis was on Feasible Region Calculus, a fundamental theoretical analysis of timing issues in distributed systems. I received my B.Tech and M.Tech degrees from the Indian Institute of Technology Madras, India, in Computer Science and Engineering.
I have over 25 publications in peer-reviewed conferences and journals and over 50 patents filed. I am a senior member of the IEEE.
Notable Awards and Distinctions:
Projects that I have led or co-led have been in the press: