Workhuman Web Assistant

I built an AI chatbot prototype designed to modernize website search through a conversational, content-aware experience. The frontend was developed in Next.js and TypeScript, while the AWS-based backend used Python and Node.js to ingest, process, and refresh website content on a recurring schedule. By combining automated knowledge base updates with OpenAI-powered responses, the prototype delivered more intuitive, accurate answers and guided users to relevant site content more effectively than traditional search.

Tools used:

Next.js, TypeScript, AWS, Python, FAISS, Puppeteer, AWS EventBridge, OpenAI

The Challenge

Traditional site search made it difficult for users to quickly find the most relevant content across blogs, landing pages, and other website resources. The goal was to create a more natural discovery experience that could answer questions directly while also guiding users to useful content across the site. A key challenge was ensuring the chatbot stayed current as website content changed frequently, including regular blog publishing and page updates, while still maintaining accurate, fast, and contextually relevant responses.

The Solution

I designed and developed a conversational chatbot experience with a Next.js and TypeScript frontend and an AWS backend powered by Python and Node.js. To keep the experience current, I implemented an automated ingestion pipeline that refreshed the knowledge base on a recurring basis as site content changed. I also integrated OpenAI GPT-4o mini to generate site-specific responses, supported by content retrieval and session context to make interactions feel more natural and useful. The result was a highly interactive prototype that reimagined search as a guided conversational experience, helping users surface answers and navigate to relevant content more efficiently.