Jun 2025 – Present · 1 yr · On-site

Jul 2020 – Jun 2025 · 4 yrs 11 mos · On-site

• AWS Sagemaker AI Apps (2022-Present)
• Developed a managed service in the cloud for running and deploying AI-based software for training and monitoring Large-language models.
• The product has been adopted by multiple large enterprises, with annual ARR of $1M+
• Led the design and development of the entire CRUD application and its APIs along with the backend infrastructure to manage the application provisioning and interaction using technologies such as AWS API-Gateway, Lambda, DynamoDB, Route53, etc.
• Scaled the service to serve over 100 requests/second into over 16 regions worldwide.
• AWS Sagemaker Studio Lab (2021-Present)
• StudioLab is one of the most popular websites out there used by students, university teachers and professionals to learn building machine learning models
• The website currently has over 350,000 active monthly users. Since its launch in 2021, StudioLab has gained continuous adoption, and is a premier product for launching and running ML notebooks in the cloud
• Developed critical components of the StudioLab website, such as the customer registration system and abuse prevention and scaled the service to support 1 million total customers
• AWS Sagemaker Studio (2020-2024)
• Sagemaker Studio is one of the most popular commercial services used by millions of users worldwide to develop and test machine learning models using its notebook services. The product has an annual revenue of over $17M+ with a 100% YoY growth.
• Developed features to increase availability of the backend service for the Studio product as part of revamping the entire service.
• Worked on decreasing overall latency of application startup times by 90% and increasing regional availability by 85% by making use of an innovative design for realtime disk backups

Feb 2020 – Jul 2020 · 5 mos · Santa Clara, C

• Part of the Amazon Alexa Geolocation service team, that provides map-based routing and storefront location and metadata
• Designed the information retrieval system to run a similarity matching algorithm based on user queries (e.g. find the “top 10 nearest hardware stores that sell nails”)
• Worked on reducing the total number of open issues in the team and led closure of over 15 high-priority customer issues

Jul 2018 – Jan 2020 · 1 yr 6 mos · On-site

• Transaction Risk Management Systems (TRMS)
• TRMS is Amazon’s fraud-detection and abuse prevention platform that prevents misuse of Amazon.com retail website services by malicious actions. The platform handles traffic for all the orders that are placed in every Amazon.com locale - which is roughly 12 million orders a day or over 360 million orders a month.
• Prevent inventory-hold abuse by sellers by integrating a real time inference model for evaluating false order cancellations. The system is part of evaluation of over 2 million orders on Amazon.in website daily.
• Led the design and development of the backend services that gather data in realtime to use as features for evaluation of customer orders
• Worked with stakeholders in multiple regions to enable integration of this new abuse prevention feature directly into Amazon.com's website order pipeline
• The new service direcltly led to over 70% reduction (estimated) of inventory-hold abuse seen on the Amazon.in website, which accounted for upto 20,000 orders a day.

Jul 2016 – Jul 2018 · 2 yrs · On-site

• Transaction Risk Management Systems
• TRMS is Amazon’s fraud-detection and abuse prevention platform that prevents misuse of Amazon.com retail website services by malicious actions. The platform handles traffic for all the orders that are placed in every Amazon.com locale - which is roughly 12 million orders a day or over 360 million orders a month.
• Designed and built a new data aggregation and feature engineering system used in machine learning models for automated investigations (e.g., Deep Neural Networks, Random Forests)
• Simplified feature variable calculation by replacing a legacy system with a new application built on an internal data aggregation platform (Java, Spring)
• The new system reduced the time required to create features by 90%, significantly increasing the velocity of new feature development
• Successfully deprecated the legacy application in favor of the new solution, improving maintainability and scalability of the feature pipeline
• Optimized backend service infrastructure usage by vertically scaling the fleet for feature calculation service which decreased overall service latency and increased performance
• Additionally, tuned the service configuration for compute to save up to 40% on infrastructure cost (approx $300k per month or $2.4million annually)
• Led development of an internal platform to support novel e-cash transactions and prevent cashback abuse on AmazonPay Wallet (India)
• Designed and implemented backend services and web interfaces for detecting and investigating duplicate customer accounts exploiting cashback offers (Java, Spring, RPC, SoA architecture)
• Built the website and backend service integrating with Amazon wallet payment services to support new e-cash transaction flows
• Developed frontend components of the investigation portal, used daily by hundreds of investigators (JavaScript, jQuery, HTML, CSS, JSP Servlets)

May 2016 – Aug 2016 · 3 mos · Remote

• Working on EcoData Retriever project (www.github.com/weecology/retriever) for NumFocus organisation under the prestigious Google Summer of Code scholarship program.

Aug 2015 – Oct 2015 · 2 mos · Remote

• Created Scilab Textbook Companion for a cryptography textbook by Prof Atul Kahate..
• http://scilab.in/textbook_companion/generate_book/3544

Jun 2013 – Aug 2013 · 2 mos · Gurgaon

• Taught a 40-hour class to middle and primary school students at the end of which they were able to develop simple sprite-based games.