May 2025 – Oct 2025 · 5 mos · San Francisco, California, United States · On-site

• Working on the EOR product

Sep 2024 – May 2025 · 8 mos · Atlanta, Georgia, United States · On-site

• Working on Sparse Attention mechanisms
• Worked on transparent performance modeling system for DL training that uses device emulation.

Jun 2019 – Sep 2024 · 5 yrs 3 mos · Karnataka, India · On-site

• Microsoft Research - Singularity project
• https://arxiv.org/pdf/2202.07848v2.pdf
• Worked on Elastic Scheduling of AI Workloads: Virtualized CUDA and NCCL driver calls, enabling transparent checkpointing and elasticity for deep learning jobs. Debugged critical low-level system issues in memory management and dynamic linking to
• enhance system reliability.
• Azure AI Platform Inrastructure
• DRI tooling: Led a team of 3 to implement API-driven automation for on-call operations, reducing unrestricted production access by 55% and enhancing IP data security with fine-grained authorization policies.
• Microsoft Research - Personalized retrieval project
• Lead the project on personalizing the retrieval layer.
• Published in SIGIR 2023 - https://dl.acm.org/doi/abs/10.1145/3539618.3591749
• Proposed the XPERT algorithm for personalizing retrieval of items from a corpus of billions of items, by introducing a novel concept of morph operators to do personalized retrieval in a resource efficient manner. XPERT offers 5% superior recall@100 than state-of-the-art techniques for personalized recommendations.
• Worked on Autosuggest for Bing project
• Azure VMware Solutions
• Designed and delivered the ‘Auto Host Remediation’ feature that monitors AVS bare-metal infrastructure and replaces unhealthy nodes automatically
• Developed the command line interface for AVS CloudSimple which was demonstrated at Microsoft’s annual conference, Ignite 2019
• Worked on SDDC buildouts – provisioning regions of VMware cloud on Azure BareMetal from scratch

May 2018 – Jul 2018 · 2 mos · Hyderabad, Telangana, India

• Developed AI-Powered Customer Feedback Analysis: Designed an AI system to provide automated insights to improve NPS and CSAT by analyzing survey responses.

Jun 2017 – Jul 2017 · 1 mo · Gurgaon, Haryana, India

• Implemented a multi object tracking system using Kalman filter of road users for ADAS applications.
• Used the YOLO (You look only once) detection algorithm, and made a model for Kalman Filter to filter out the errors in the detection algorithm and create associations in the consecutive frames at the same time. Achieved real time performance in tracking.

Dec 2016 – Jan 2017 · 1 mo · India

• Plagiarism checker for toy assembly language SimpleRisc.
• The project involved designing and implementing an algorithm to detect plagiarism with
• simpleRISC assembly codes. SimpleRISC is a toy assembly language with a limited instruction set.
• The algorithm designed included breaking the code into basic blocks then to string, using levenshtein distance for string matching, designing a proper matrix for plagiarism.