Car for Jane

Inspiration: At my daughter’s school, nurses and moms kept telling me the same story: “We just moved, I’m starting a new job, and now I need a car — I’m overwhelmed.”

They weren’t asking for options. That’s why we built this — to offer support, a human-like chat that gives a clear answer to what car is for Mom to commute to her new job.

What it does:Meet Jane — a nurse, mom of two, and the reason we built this app. After too many stressful car-buying experiences, she didn’t need more filters — she needed peace of mind. Especially now, freshly relocated to San Ramon, juggling Bay Area life and a new job, everything feels like too much.

Our AI chatbot meets her like a friend — not with endless options, but with clear, caring answers. It gives her just 2–3 car choices, personalized to her needs, her budget, and in her language. No pressure. No price up. Just a simple, honest conversation that ends with a confident decision.

How we built it

We used React Native (with Expo) and TypeScript to keep the app fast, mobile-first, and clean. Supabase powers our backend with a real-time PostgreSQL database, Edge Functions for logic, and Row-Level Security for safe user data.

The AI car-matching is built on a RAG pipeline using Gemini and Claude, and we developed everything inside Cursor AI IDE to move quickly while keeping the stack connected. Blockchain adds traceability for future trust layers without overcomplicating today’s build.

Challenges we ran into

Designing for someone like Jane — who’s not just busy, but emotionally overwhelmed — meant we had to rethink everything. The hardest part was making the chatbot feel like a human conversation, not a form with friendly words. We also had to simplify complex decisions like financing and insurance without losing trust or clarity.

Accomplishments that we're proud of

We built something that gives Jane real answers — not options — in under 30 seconds. The app works offline, supports English and Spanish, and never overwhelms. Most of all, it feels like someone on her side, not another app asking her to figure things out.

What we learned

People don’t want more features — they want clarity and care. Designing for someone who’s overwhelmed taught us to slow down, listen deeper, and build with empathy. Simplicity isn’t basic — it’s powerful.

What's next for Car for Jane

We’ll start with Supabase and pgvector, but we’re open to expanding to LangChain or LlamaIndex as data grows. Gemini is our preferred choice for context-rich responses, and we may add fine-tuning later for better personalization. Future updates could include EV incentive matching, insurance estimates, voice chatbot, and trusted local dealer partnerships.

Built With

• and-developed-inside-cursor-ai-ide.-we-also-use-supabase-edge-functions-for-serverless-logic
• and-plan-to-expand-with-langchain
• and-supabase-for-the-backend-with-postgresql-and-pgvector.-the-ai-layer-runs-on-a-rag-pipeline-supported-by-gemini
• blockchain
• claude
• cursor
• deepseek
• gemini
• llamaindex
• python
• rag
• reactnavitewithexpo
• supabase
• typescript