Projects

I developed an AI agent using Retrieval-Augmented Generation (RAG) to streamline product support by answering queries based on documentation from multiple sources, including Contentful, GitHub, and Notion. Key contributions include:

• Building a Node.js API layer to crawl the knowledge source and return data in a uniform markdown format.
• Building the workflow pipeline that calls the API, generates content chunks using LangChain, converts them into embeddings using AWS Bedrock and lastly stores the embeddings into a Vector Database (Pinecone).
• Building an API endpoint to retrieve, augment, and stream AI-generated answers in real-time to the frontend.

The solution uses AWS Bedrock for both embedding creation and answer generation.

I built the developer portal for Tray.ai from scratch in just 1 month as a single person dev team. It was built using Redocly allowing writers to write content in plain Markdown files. Key contributions include:

• Designing and developing the landing page for the portal.
• Creating comprehensive OpenAPI specifications for all the Tray.io APIs.
• Building essential developer tools to facilitate onboarding of API customers.
• Developing custom components using React to enhance the functionality of the portal.
• Implementing an automated pipeline to populate release notes from the product team.

This project provided a centralized resource for Tray developers to access API documentation, tools, and updates, significantly improving the developer experience.