Jan 2023 – Jan 2024 · 1 yr · San Francisco Bay Area

• Recruited as a senior technical leader to solve high-priority challenges in Meta's core Ads Ranking system. I was responsible for architecting the strategy for model evaluation, and defining and improving the core metrics used to validate and ship next-generation deep learning models.
• I left to go on a startup journey, as building in GenAI is opportunity not to be missed at this time!

Mar 2022 – Aug 2023 · 1 yr 5 mos · San Francisco Bay Area

• I led the machine learning team that built the core recommender systems for Uber's Growth & Marketing division. My team's work focused on advancing personalization for Uber and UberEats by innovating in retrieval and ranking models, building scalable vector search infrastructure, and applying graph algorithms (like PageRank) to enhance recommendation relevancy. These systems power our personalized email/push notifications and creative optimization engines.
• In addition to my role as Tech Lead, I served on the "Ark Committee," a select group of ~30 senior technical leaders shaping Uber's company-wide technology roadmap and standards.
• Read more about my team's approach to personalization: https://www.uber.com/blog/personalized-marketing-at-scale/

Dec 2020 – Feb 2022 · 1 yr 2 mos · San Francisco Bay Area

• As the Machine Learning Tech Lead for Pinterest's shopping discovery, I led the team responsible for ranking and retrieving content across all major surfaces, including Homefeed, Search, Related Pins, and the Shop Tab. My work delivered personalized shopping experiences to over 400 million users by navigating a corpus of 1 billion+ items.
• My leadership spanned the full ML stack, from deep learning-based retrieval and fast vector search to multi-objective optimization (MOO) and user intent understanding.
• Published Work (Solution for streaming-enabled fast vector search): https://medium.com/pinterest-engineering/manas-hnsw-streaming-filters-351adf9ac1c4

Jan 2018 – Present · 8 yrs 2 mos

• Investor/Advisor
• GOAT Brand Labs, Ananant Systems, Goswirl.live, Metaphor, Mage.ai, Chroma Inc, Fennel.ai, Nected.ai, Observe.ai, Beacons.ai, Proshort.ai, Databrain, Revsure.ai, Middleware, Innerfit, Recur Club, Republic, DoctorC, Causal, Farmstead, Verifiable, Quattr, Spike.Club, Edureka(Acquired by Veranda Learning), EzCred(Acquihire by Coinbase), Helo (Acquired by MyGate)
• LP
• Emergent Ventures, Zetta Ventures, Essence Venture Capital

Sep 2015 – Jan 2021 · 5 yrs 4 mos · San Francisco Bay Area

• As a Staff Engineer, I was responsible for the end-to-end representation learning and interest based recommendations for Twitter's Home feed and Ads systems, spanning from deep learning modeling to retrieval infrastructure.
• ML Architect (Representation Learning): Led the technical strategy for Ads Ranking and Cortex ML teams, designing the deep learning models (Two-Tower DNNs) used to learn user and tweet representations for Homefeed and Ads.
• Vector Search & Retrieval: To power this new model class at scale, I authored Twitter's in-house vector database (open-sourced here) to enable fast, real-time retrieval.
• Foundational: Co-authored with Dr. Yury Malkov, an algorithm for dynamic HNSW updates, which was later open-sourced and merged into the industry-standard hnswlib library and being leveraged across the industry for vector Search.
• Strategic Initiative: Selected by the CEO (Jack Dorsey) and his staff for a confidential, high-priority project to drive the next wave of user growth in developing markets
• Publications
• KDD'2020: Two Tower DNN based tweet retrieval (https://arxiv.org/abs/2105.09293)
• Twitter Recsys’2020: Organizer and Reviewer
• https://recommender-systems.com/news/2020/11/09/what-twitter-learned-from-the-recsys-2020-challenge-michael-bronstein-et-al/
• Embeddings@Twitter
• https://blog.twitter.com/engineering/en_us/topics/insights/2018/embeddingsattwitter.html
• OSS:
• With Dr. Yury Malkov: Dynamic Updates For HNSW - Hierarchical Navigable Small World Graphs. Code open sourced as part of hnswlib (Fastest Nearest Neighbor Search in the industry). https://github.com/nmslib/hnswlib/pull/216

Aug 2014 – Sep 2015 · 1 yr 1 mo · Bangalore,India

• Acquired by Walmart
• (https://www.recode.net/2018/5/11/17335708/walmart-flipkart-india-ecommerce-charts)
• Worked on e2e design and implementation of first mile logistics and compliance system for logistics
• Design and development of sense and response platform that ingest events across the logistics and provide tools to perform batch and real time analytics on top of it.
• Android app for first mile marketplace pickup for warehouse and delivery people (Interest Project)

Sep 2012 – Aug 2014 · 1 yr 11 mos · Bengaluru, Karnataka, India

• One of the early engineers for the first deep tech development office of WalmartLabs in India.
• Design and development contribution to the Marketplace platform for world's largest retailer Walmart.
• Worked on building and scaling Marketplace platform.
• Contribution in building and designing the notifications infrastructure for transactional Email/SMS/Push notifications.

May 2011 – Sep 2012 · 1 yr 4 mos · Gurgaon, India

• Proud to be the early engineer for the first Indian E-Commerce company to go IPO.
• (https://www.vccircle.com/infibeam-open-first-ipo-indian-e-com-venture-march-21/)
• Designed and developed multiple core components from scratch :
• Order management system
• Fulfilment system
• Product CMS
• Warehousing system