Oct 2022 – Jul 2025 · 2 yrs 9 mos · Redmond, Washington, United States

• Working on Azure Storage's Stream Layer, a distributed file system powering Microsoft's cloud storage infrastructure.
• Key Contributions:
• Led critical improvements to data durability and availability mechanisms in distributed storage
• Architected performance optimizations reducing network traffic and improving resource utilization
• Developed system protection mechanisms for high-scale distributed services
• Enhanced observability and monitoring capabilities, significantly reducing incident resolution time
• Led team initiatives in performance testing and system stability

Jan 2021 – Oct 2022 · 1 yr 9 mos · Redmond, Washington, United States

• Working on Azure Portal's low-code/no-code platform, enabling developers to build and deploy experiences across Microsoft's cloud services.
• Key Contributions:
• Designed and implemented configuration management system for Portal extensions
• Built reusable component infrastructure serving thousands of daily users
• Enhanced developer productivity through automated tooling and frameworks

May 2019 – Jul 2019 · 2 mos · Hyderabad Area, India

• Worked on ​ Post-Copy of Resource disk ​ in a Live Migration scenario:
• Performed ​ Proof-of-Concept ​ of post-copy based migration of resource disk, a previously unexplored territory in Azure/Hyper-V VM migration landscape.
• Business Impact ​: It will help reduce VM downtime due to
• host-maintenance/degraded-hardware, ​ from minutes/hours to the order of seconds ​.

May 2018 – Jul 2018 · 2 mos · Montreal, Canada Area

• Mining Twitter for Developers
• ◦ Conducted an exploratory study comparing the effectiveness of machine learning classifiers and deep learning-based models in classifying tweets that can be resourceful to developers.
• ◦ Showed that attention-based LSTM models and the ones leveraging sentence/ word level transfer of knowledge outperform Naive Bayes and SVM, even in a paucity of training data.

May 2017 – Jul 2017 · 2 mos · IIT Kharagpur

• Dynamic Design Discovery of Thread Models:
• ◦ Modeled deep learning-based classifiers fed by run-time events traced according to pre-specified logic to infer thread models like concurrency, data access, exception, etc. of multi-threaded applications.