Sep 2021 – Present · 4 yrs 6 mos

• I took over as the overall Engineering and Product lead for Knowledge Cards, with the goal to scale to 100+ languages in the world, along with a bottoms-up implementation for mobile. Working closely with my team in China of 10 Engineers, and coordinating across both time-zones, we were able to exceed the goals set. I also had the opportunity to be recruited into a selective Bing-wide effort to integrate the power of Large Language Models into Search. My team drove and pioneered a few key efforts with this technology - which we called 'Stories' and 'Knowledge Cards v2'.
• I was also deeply involved in the release of Bing Chat. I helped refine the search grounding framework, drove the initial implementation of the authentication protocols and helped integrate location signals into the response construct of Bing Chat, enabling answering location based queries with reasonable accuracy. Over the next few months, my team also drove other efforts on Bing Chat, including the Creator, Chat, Companion answers and enabling Knowledge Card images on Bing Chat. I was also given responsibility to drive the caching infrastructure for Bing Chat to minimize GPU costs on higher frequency queries. My team was able to engineer a solution that drove 25% GPU savings.
• By this point, my team had grown considerably, to about 25 engineering in China and 30 engineers in USA. I was also helping psuedo-manage design teams and PM teams alike, to achieve our new goal - to reconstruct the entirety of the Search Engine Results Page. One key deliverable here was the 'Magazine' project, which allowed us to present broad queries with a grid UX on top of SERP showing various mixed intents. We also embarked on the boldest reimagination of Search yet, by revamping the entire look and feel of Search Results pages by leveraging the power of Generative AI. I drove the vision, architecture, direction and implementation of this project end-to-end and it was a startup-style project that was driven internally.

Nov 2019 – Sep 2021 · 1 yr 10 mos

• In this phase of my career, I transitioned to taking ownership of the Bing Recipes answer. This was an end-to-end ownership including owning the data pipelines, the search index, the storage layer, the runtime logic and the UX. I started off as the tech lead for the project on a team of 7 engineers including myself and shortly transitioned to become the Engineering Manager of the group as well, with the other 6 reporting directly to me. I was able to introduce new paradigms our team was doing differently from other teams - regular planning, team retrospectives and a strong culture of actively supporting success for each other.
• In my tenure as owner of Bing Recipes, we achieved several remarkable milestones for Bing search. We doubled our daily Bing Recipes query volume and increased our DAU by 66%. I drove the re-architecture of the entire data pipeline and index design to support becoming a fully universal answer, ready to launch in 100 languages and 236 regions worldwide. In addition, we also had several new milestones for Bing answers - we were the first team to introduce videos on the experience and the first team to add filters personalized with user and query signals. Unfortunately, due to a major re-org in Bing, we had to hand over the reins of the project at the end of a year.
• Following this, I embarked on my most ambitious project yet, the Bing Knowledge Cards effort, which was a complete redesign of the traditional entity pane experience familiar to all in Search. My primary role was to take an early, manual effort, and achieve the very difficult goal of building 250,000 cards at scale in 6 months. I worked with a team of 5 direct reports, 4 dotted line reports from around the world and about 40 vendors located worldwide. We successfully laid clear goals and strong timelines on how we wanted to achieve this, and I drove the technical design as well as the product goals, timelines and execution in an agile manner to achieve the tough goal that was set.

Sep 2018 – Nov 2019 · 1 yr 2 mos

• I was a Senior Engineer on the Enterprise Knowledge Graph team, later known as Microsoft Enterprise Graph. We built a product that leveraged the technologies built in Bing's powerful Knowledge Graph and apply it to Enterprise data, allowing efficient storage, entity linking and powerful search capabilities for enterprises, including the ability to search with free-flowing multi-turn conversation.
• My role on this team was multi-fold - I helped build multiple customer demos that were presented at Microsoft Ignite, Microsoft Build and other conferences. I was also the key dev on the Access Control and Proxy layer of the graph, which allowed me to connect with all the other services in the team and enable service-specific authentication, authorization, API access, custom routing, load balancing and bot detection in a clean, horizontal way. I also helped build the algorithms that enabled Access Control rules at a node level on the graph data.
• Microsoft Enterprise Graph was released in a preview mode on Azure Marketplace in 2019.

Aug 2016 – Sep 2018 · 2 yrs 1 mo

• I worked on the initial implementation of the Bing app for iOS, which took a fresh take on the User Experience for mobile search and pioneered many new experiences for mobile search. My role on the initial app was to help work on image and video search verticals and I worked with a team to help implement the initial version of the code in swift using clean design patterns and an extensible design.
• I was a strong contributor, helped defined the direction of the initial iteration of the product and first hacker/engineer on the ground breaking accessibility project, Seeing AI - the talking camera app for the blind community. This project was conceptualized and initially executed at Microsoft's OneWeek hackathon in 2015, where my team won 3rd place among 13,000+ hackers.
• I also was the lead for Project DealWatcher - a Microsoft Garage app that would crawl deals on the web in near real time and surface them on a dedicated app that would allow the user to track deals. This app was implemented in Xamarin to serve Android and iOS users, with servers hosted on Microsoft Azure. I helped drive the architecture of the product, full-stack engineering including writing code and the day-to-day technical leading of the team. This helped conceptualize the Bing Coupons and Deals segment later on.
• I was part of the 3D Search team on Bing. 3D search had many surfaces like enabling search in a 3D environment like Augmented Reality on HoloLens and Virtual Reality headsets. I worked on building Augmented Reality solutions on Android, iOS, HoloLens and worked on the 3D SDK for Web that allowed partners to integrate the 3D viewer on their web sites. We also used this enable 3D ads for Samsung during their launch of the Galaxy S10 Phone.

Oct 2014 – Aug 2016 · 1 yr 10 mos

• I joined the Bing Search History backend team as an entry level Software Engineer. The Bing Search History team handles high volumes (~100,000 queries/sec) data reads and writes and has immense impact on various partners across Bing, including Autosuggest, search timelines and personalization scenarios. My role on this team was to initiate and execute a migration from a legacy database to newer, more centralized key-value store within Microsoft. Along the way, we had to migrate code across various languages, build offline pipelines to migrate the existing data, wire up writes to both stores and monitor the processes till we had 100% parity between both systems. I assisted in all these aspects and primarily handled the reads and writes from the new store, ensuring scale, code quality and feature parity was executed at the highest quality.
• For the second half of my year in the project, I was a primary engineer on a project to scrape mobile apps on Android and iOS and build a search index using scraped app data. This required me to learn the internals of both operating systems, build a framework to seamlessly scrape various data from both operating systems and build a cluster of on-demand mobile devices of various resolutions to be able to run this at scale.