Dec 2021 – Present · 4 yrs 3 mos

• I work with a team of Applied Scientists and Machine Learning Engineers on retrieval and ranking for Sponsored Products(any ad you see with the word "Sponsored" on Amazon). We support the business by developing low latency systems to search more than 100 billion ads for more than a billion Amazon products in order to select a few hundred ad candidates based on shopper journey. I lead work on vector and text based hybrid retrieval using vector databases storing high dimensional dense embeddings from sentence transformer based bi-encoder models and ranking using cross-encoder models. Developed services and features to expand coverage of automatic and negative targeting features provided by Sponsored Products. Also led Spark infra migration to Kubernetes clusters running ETL and inference on 1000+ nodes.
• Currently working on developing agentic frameworks to update advertiser campaign targeting and bidding.
• https://advertising.amazon.com/solutions/products/sponsored-products

Aug 2020 – Dec 2021 · 1 yr 4 mos

• Prime Video - Content recommendations for the top banner and the rows below you see on Prime Video homepage. Real-time machine learning inference to select best title based on customer and title data.

Jul 2017 – Jul 2020 · 3 yrs

• Ad content moderation - To make sure ads coming up on Amazon website don't violate Amazon's advertiser policies. Worked on machine learning backed automation to moderate all ads seen and worked with ML scientists to check more than 10 million ads daily and ensure our customers are not exposed to ads violating the advertising policies.