Lab
Some Experiments and work in progress
B-Bot is a React-based frontend chat interface powered by a FastAPI backend with Retrieval-Augmented Generation (RAG). The assistant, based on HuggingFace's zephyr-7b-beta, uses semantic search on embedded knowledge documents to provide grounded responses in my tone-of-voice, a UX strategist and AI product manager. (in training)
User → Chat UI → Cloudflare → FastAPI → FAISS → FastAPI → HF API → FastAPI → UI
🧠 Features
Chat interface mimicking my tone of voice/personality
Retrieval-Augmented Generation (RAG)
Document embedding and semantic search
Real-time chat experience with loading states
Persistent session using browser local storage
🧭 User Stories
1. As a user, I want to ask B-Bot questions and receive helpful, Bernadette-style answers.
2. As a developer, I want to embed documents and regenerate a vector index.
3. As a UX researcher, I want to trace if the bot is using my context or hallucinating.
4. As a product manager, I want strict instruction-following behavior to ensure grounded responses.
Stack Overview
Frontend
React (Framer Code components, Inline CSS)
API Proxy
Cloudflare Workers
Backend
FastAPI (Python 3.12)
Embedding Model
SentenceTransformers
(all-MiniLM-L6-v2)
Vector Store
FAISS
Deployment
GitHub + Codespaces + Cloudflare
From UX to AI:
Mapping a Journey of Innovation
(in Tableau)
Explore this dashboard to discover how I’ve contributed to AI-powered product ecosystems—from IBM Watson and TD's contact center automation to research repositories and chatbot solutions.
These aren’t just cool projects—they’re blueprints for AI done right.
