Dec 2025 – Present · 3 mos · Coimbatore, Tamil Nadu, India · On-site

Aug 2023 – Nov 2025 · 2 yrs 3 mos · Edmonton, Alberta

• Took time to travel, explore startup ideas, learn early-stage venture fundamentals, and support personal commitments and transitions.

Jan 2021 – May 2023 · 2 yrs 4 mos

• Microsoft Teams Engineering
• As the tech lead, I helped Teams App increase its scalability to over 5x on its channels' vertical. This process began with developing automated internal tools to assess the feasibility of scaling while ensuring optimal app performance. After analyzing the data gathered and obtaining consensus through design reviews, I worked on scaling across the stack including client-level APIs, in-memory data layers, backend data layers, and backend APIs. I facilitated collaboration with over ten teams (across time zones) responsible for critical modules within the Team's App, ensuring timely delivery of this feature to our customers.

Sep 2020 – Dec 2020 · 3 mos

• Bing Client Connectivity

Jun 2019 – Aug 2020 · 1 yr 2 mos

• Bing LiveSite Engineering

Dec 2017 – May 2019 · 1 yr 5 mos

• I could not have asked for a better start to my job life! In this position I worked with Bing Live Site Engineering Team. My responsibilities broadly included monitoring Bing's Reachability from across the Globe, improving Bing and it's sister teams' resiliency to DNS related outages and improving Bing's DNS - Config as Code architecture. It has been a ride and of course Vancouver is the best :)

Jul 2017 – Jul 2017 · 0 mo · Bengaluru Area, India

• We started with the problem of anomaly detection amongst various clinical samples based on their wet-lab and bioinformatics attributes. We approached this problem using two different techniques - (1) Principal Component Analysis followed by K-Means Clustering, and (2) Box-Plot and Outlier Detection using Z-Scores. Both these approaches gave the expected results. Later we worked on a classification problem of identifying the tissue of origin from the somatic mutations of a tumor specimen. We used a Random Forest Model as suggested by a previously published work. All the machine learning code was written using the Language R - along with the support of scripting languages like Bash and Awk and also C++ for the required data processing.

Jan 2017 – May 2017 · 4 mos · Kanpur Area, India

• ESC101 (Fundamentals of Computing) Tutor, under Prof. Rajat Mittal

May 2016 – Jul 2016 · 2 mos · Melbourne, Australia

• Enhancing a given SAT encoding to become a propagation complete encoding (PCE) by adding additional clauses of pruning information for unit propagation is an important and interesting problem. Though the complexity of the algorithm we present is exponential (which has been achieved in the past as well), it is more intuitive, simple and works with a reasonable efficiency for a practical pre-processing of encodings. We use the concept of Minimal Unsatisfiable Cores (or MUSes) for our pruning while we explore all the possible partial assignments of the variables. The implementation has been done in Haskell using the support of Picosat Haskell Package.
• Supervisors: Assoc. Prof. Harald Sondergaard, Dr Graeme Gange

May 2015 – Jul 2015 · 2 mos · India

• BEST SHOT EXTRACTION FROM NEAR DUPLICATES
• Feature Extraction was performed on input images based on content and quality of the photo.
• Dlib-ml toolkit was heavily used in our implementation.
• Several Machine learning based models like SVM, Decision Trees were trained on these features.
• Our results showed that Random Forest model is very promising.