🚀 Built the experimentation infrastructure powering Rufus's growth

Redesigned Amazon Rufus's A/B testing platform from scratch, enabling 12+ teams to run concurrent experiments on the same product or keyword simultaneously — something that was previously impossible. Experiments on this platform drove +40% growth in Rufus monthly active users and +$12.74M in sponsored product revenue.

• --

🤖 Shipped an LLM-powered AI agent for Amazon's shopping assistant

Led end-to-end delivery of a Claude-powered agent that generates contextual follow-up questions for Rufus users at 10 requests/second. I designed a single unified agent capable of handling two distinct page contexts with different evidence sources, hit the peak shopping season deadline, and achieved a 69% quality win rate over the previous system.

• --

⚡ Cut ML inference latency by 50% — without retraining the model

Identified that Alexa's answer ranking model was being called 30 times per query instead of once. Restructured the invocation pattern to batch all candidates into a single call on AWS Inferentia hardware. Capacity grew from 4 to 30+ answers per query, latency dropped by 50%, and voice-shopping answer coverage grew ~40% in the US — all delivered in under 3 weeks.

• --

🔗 Built the content injection framework that powers Rufus's dynamic features

Designed the first-ever framework in Rufus for blending non-AI-generated content — such as customer reviews, deals, and pricing — into the AI assistant's response stream. The reusable patterns I established were adopted by 6+ subsequent feature teams, eliminating the need for architectural rework on every new content type.

• Designed architecture and implemented java microservices for the company’s help website called Intelligent Help Interface (IHI) which is being highly adopted by the end users to access the efficient help solutions.

• I worked on creation of the REST API’s using spring boot, integration with Salesforce, QR Code generation, SQL utilization, AWS services, cache management and Kubernetes deployment.

• Leading the team of 6-8 developers for the overall project and co-delivery owner for IHI, which includes business co-ordination, security reviews, monitoring, sprint planning, integration with game teams, project demos and quarterly planning.

• Worked as an individual contributor to design and architect the backend of the GCN application which facilitated EA portfolio game influencers for efficient payment transactions, content upload, etc. at high scale.

• This project involved me in the implementation of REST API’S, PayPal integration, AWS Services, Security concerns, Exception handling, logging, Testing framework, Configuration management, etc. from the ground up.

• Worked extensively on docker containers & AWS services like EC2, ECS, ECR, CloudWatch, CloudFront, S3, SQS, API Gateway, etc.

• Contributed as a Full-Stack Engineer for a metric analysis software tool for Competitive Gaming Division.

Working on the development of OZO-virtual reality camera using NodeJs, ReactJs, HTML, CSS, Scala and Play Framework.

The development involves fetching of data using REST API’s and Web Sockets and displaying the processed data on the web page.

Also developed a web application using electron, nodejs and react for implementation of node discovery feature.

Working as a Full Stack engineer for the engineering team.

Working on AngularJs, Javascript, HTML and CSS to develop an internal website named columbus.

Also working on Scala for the backend programming of the website.