Jul 2024 – Sep 2025 · 1 yr 2 mos

• As the director in SageMaker AI group, my teams build state of the art distributed training technology to enable customers to train, finetune and host foundational models. SageMaker AI is an industry leading managed service for building with ML and GenAI and is used by over 100K AWS customers. We offer a fully managed infrastructure, distributed training libraries and training recipes for large-scale and cost-effective FM training via SageMaker HyperPod and SageMaker Training jobs on thousands of accelerators (GPUs and Tranium). These services are used by leading startups and enterprises to train or finetune the state of the art Foundation models (e.g. Amazon Nova, Stability.ai, Perplexity.ai, Luma.ai).

Nov 2021 – Jul 2024 · 2 yrs 8 mos

• As the GM of SageMaker's LCNC (Low Code No Code) group, I led a multi functional team of engineers, researchers, product managers, UX designers to build a portfolio of low code/no-code products of SageMaker (e.g. SageMaker Canvas, SageMaker Jumpstart, SageMaker Data Wrangler, SageMaker Automated Model tuning). My team was responsible for democratizing the use of Gen AI and ML by making it easier for customers to use this technology to customize/finetune models (no code RAG, AI agents for data preparation and model building).
• 1. No code ML: Sagemaker Canvas that allows customers to build and use GenAI and ML models without writing any line of code.
• 2. Low code ML: Services that reduce undifferentiated heavy lifting for expert practitioners. Offerings include: SageMaker Jumpstart and algorithms (state of the art models and solution repository), SageMaker Data Wrangler (a low code data prep tool), SageMaker AutoPilot (AutoML for training ML models), SageMaker AMT (Automated model tuning for hyper parameter optimization).

Jan 2020 – Nov 2021 · 1 yr 10 mos

• As the director of engineering, I led engineering teams for AWS Computer Vision services group. My teams were responsible for AWS computer vision services e.g. Amazon Rekognition, Amazon Textract and Amazon Lookout for Vision. These services are used by our customers to analyze billions of images, extract information from millions of documents and analyze thousands of hours of video every week. Our vision was to make computer vision technology accessible to every developer and a business user. We do it by offering managed services that customers can call using APIs or UI.
• These services enabled customers to
• Quickly add pre-trained or customizable computer vision models to applications without building machine learning (ML) models and infrastructure from scratch.
• Automatically extract printed text, handwriting, and data from any document.
• Spot product defects using computer vision to automate quality inspection.

Sep 2015 – Jan 2020 · 4 yrs 4 mos

• As the director of core shopping, I led the product and engineering teams responsible for Amazon's retail shopping website and shopping apps (esp. Amazon's Homepage, Product Detail Page, Navigation, Shopping Cart and Checkout). These applications are used by millions of customers every day to shop on Amazon. I joined this group in 2009 and had opportunity to lead several critical initiatives for the company. Examples include: modernizing tech stack by developing platform that powered Amazon's retail website and shopping apps using the same code base, radically improving latency for customers, building content management systems. In addition to platform and tech initiatives, my teams also delivered $20B+ in incremental revenue by modernizing customer experience. This required us to try several bold bets and run thousands of experiments at scale. My teams were cross functional, geographically distributed (Seattle, Austin, London and Bangalore) and comprised of product managers, software development managers, software engineers, applied scientists and business intelligence engineers.

Oct 2013 – Sep 2015 · 1 yr 11 mos

• I led the team that was responsible for Product Detail Pages of Amazon.com across all market places (North america, Europe and Asia) and owned product innovation, tech platform, operations and latency improvements. Millions of customers around the world shop from Amazon's product detail pages from just about every device imaginable and my team's responsibility was to innovate on behalf of our customers to make Amazon's product pages blazingly fast, define the architectures and processes to enable our continued growth and success, drive technical strategy across multiple teams while maintaining high availability, growing revenues, and reducing costs.

Jul 2009 – Sep 2013 · 4 yrs 2 mos

• Responsible for product detail pages for Amazon.com and all marketplaces.

Jan 2007 – Jul 2009 · 2 yrs 6 mos

• I was the technical delivery manager for the commission processing system that was used to process commsion for insurance agents. I build the product and owned the the first product deployment at MetLife Financials USA. During this time, i was also the product panager for Multi Channel Commerce and Point of Sales Application for Prudential Insurance.

Jan 2005 – Dec 2006 · 1 yr 11 mos

• I was the tech lead for order management system for Orange Business telecom.

Jun 2002 – Dec 2004 · 2 yrs 6 mos

• Responsible for SGI's Storage and networking solutions configurator development.

May 2001 – Jul 2001 · 2 mos · Paris, Île-de-France, France