Zenith AI

Inspiration

I’ve always wanted an AI that could genuinely support my mental health, something I could turn to anytime, ask anything, and receive a response that actually resonates.

After trying generic chatbots like ChatGPT and Gemini, I realized they often miss the depth, empathy, and personalization that mental health conversations require. That gap lit the spark for Zoe: an AI designed not just to reply, but to truly connect and help people feel seen.

What it does

Zoe listens first, asking the right questions to fully understand the user’s situation. Then, it analyzes the details and delivers a solution that fits them personally. Whether someone feels insecure, sad, unhappy, or unconfident, Zoe adapts, because the right answer isn’t the same for everyone.

How I built it

I started with a basic RAG system and grew it step by step into something powerful: Qdrant as the vector store for knowledge management. LangChain + FastAPI for tool orchestration and APIs. React + TypeScript (TSX) for a clean, responsive UI.

Over time, I layered in multiple tools and fine-tuned the way Zoe integrates knowledge, so it doesn’t just retrieve information, it uses it in a way that feels natural and actionable.

Challenges we ran into

RAG was initially slow retrieving data took too long. Tool orchestration with GPT-oss 120b sometimes caused errors when multiple tools were called together.

These pushed me to rethink architecture, optimize retrieval, and refine tool coordination until Zoe became smooth and reliable.

Accomplishments that we're proud of

Zoe now has a rich, evolving knowledge base built from books and external sources. It can blend this knowledge with real-time insights to give personalized answers that actually make sense. The experience of talking to Zoe feels closer to chatting with a supportive friend than a chatbot.

What i learned

How to build better RAG systems and make LLMs use retrieved data effectively. Deep insights into tool calling with LangChain knowing when, how, and why to integrate tools without overloading the system. The importance of human-like tone in AI, especially in sensitive areas like mental health.

What's next for Zenith AI

Mobile App for easy access anywhere. Real-time speaking capabilities so interactions feel natural and fluid. Expanded knowledge across even more mental health and personal growth resources. Smarter, faster tools for instant, reliable responses.

Built With

• fastapi
• groq
• langchain
• python
• react
• typescript