May 2021 – Oct 2025 · 4 yrs 5 mos

• Managed a team of 7+ Eng for building the visual-inspiration shopping experience from 0->1 on Google Search, including products like "Vision Match" in AI Mode (for Outfit and Room inspiration), “Shop the Look” (presented Shoppable looks relevant to the search query), "Style Ideas" (inspiration for styling a given apparel) and “Trending Styles”.
• This involved solving difficult problems that had not been done before at Google e.g.
• Scraped a corpus of 200M inspirational images from unstructured webpages, leveraging heuristics and traditional ML models and extracted fashion trends from Fashion blogs using LLMs
• Retrieved the relevant looks given an Apparel item image and algorithmically generating visually similar products) by leveraging the cutting edge image ML models / infra at Google (e.g. object detection and segmentation, image embeddings and ScaNN for vector similarity search)

Oct 2020 – May 2021 · 7 mos

• TL-ed a team of 4 eng to build a prototype for a new destination for Brands to showcase their products / lifestyle content in < 2 months
• TL-ed the design and implementation for the distribution of Shoppable short-form videos from Brands across Google surfaces

Oct 2018 – Oct 2020 · 2 yrs

• TL-ed a team of 4 Eng for building the unified index containing the image metadata and signals used for answering queries on Lens across all verticals. This significantly reduced the cost of maintaining different indices and increased the velocity of quality improvements across Lens
• TL-ed a team of 5 Eng for launching the first inspirational Shopping feature on Lens called “Style Ideas” from 0->1 (inspiration for styling a given apparel) which got very good user engagement. This feature was later also launched on Google Search

Dec 2012 – Oct 2018 · 5 yrs 10 mos

• TL-ed the Ads Eng team in Google Play in its very early days, and generated $XXXM ARR in ~1.5 years. We launched a sponsored cluster of App Ads on the Play Store homepage and later extended it to the Search results page, Games homepage and Apps detail pages. This was a pivotal change as it provided a modular and reusable framework for showing Ads anywhere on the Play store (vs building one-off solutions before) and significantly increased the velocity for experimenting with new Ad formats
• Built a recommendation system for personalizing the Apps shown on the homepage. I also launched personalized ranking of clusters on the homepage by training a new ML model. This work significantly increased the App install rates as users were able to find more relevant suggestions.
• Initiated and successfully launched separate Navigation tabs for the different verticals of the Play store thereby improving the Information Architecture and significantly increasing the traffic to the homepages
• TL-ed the migration of the “cross-sell” clusters (e.g. “You might also like”, “Similar Apps”) on the Apps detail pages into a single stream of clusters which unlocked 6 new launches by other teams. Led the server-side design and implementation for the revamp of the details page which introduced a new “look and feel” and interaction model.

Jul 2010 – Dec 2012 · 2 yrs 5 mos

• Extended SQL syntax by introducing a new kind of join to succinctly express hierarchical (cube-like) queries and implemented a backend for executing queries using the new syntax.
• Built a dashboard for reporting charitable contributions made by Google and other reporting solutions from ground-up
• Added many new features and improved performance of FlumeETL - a framework built on top of map-reduce that allows developers to write ETL (Extract Transform Load) pipelines using SQL-like operations