Bestie

Monthly Growth
Persona
Simulator
Insights

Inspiration

Managing mental health, physical wellness, and finances usually happens in silos — check-in apps, wearables, and banking dashboards rarely connect. We wanted to build a single trusted “best friend” AI that understands you holistically, nudges you gently, and provides actionable suggestions.

What it does

Bestie is a unified dashboard + AI assistant that: Pulls in your daily check-ins (mood, stress, sleep). Connects to health data (steps, heart rate, VO₂, sleep). Generates mock Plaid-like finance data to simulate real-world usage. Runs an AI-powered insights engine that summarizes your state, highlights trends, and suggests concrete steps across health, finance, and mental well-being. Provides a forecast simulator: tweak lifestyle habits (e.g., +2000 steps/day, +0.5h sleep) and visualize projected outcomes. Personalizes guidance via a persona model that evolves weekly from your behavior.

How we built it

Backend: FastAPI + SQLAlchemy + SQLite/Postgres for persistence. Frontend: React (Vite) with Recharts for visualizations and Tailwind for styling. AI: Integrated with an open LLM server (vLLM / LM Studio) to generate summaries, suggestions, and forecasts. Mock data generators for check-ins, health, and finance to ensure full demo without external APIs. Persona engine: Updates weekly to personalize tone and recommendations.

Challenges we ran into

Data integration: Making health, finance, and check-ins align on a daily timeline. AI context size limits: Models struggled with long histories, forcing us to compress data intelligently. Forecast realism: Balancing statistical projections with natural-language AI explanations. UX clarity: Avoiding overwhelming users with data while keeping the dashboard actionable.

Accomplishments that we're proud of

A working full-stack prototype that unifies health, finance, and wellness. Interactive simulator with scenario-based forecasting. Mock Plaid integration that feels real for demo purposes. A persona-driven AI coach that feels personal rather than generic.

What we learned

LLMs excel at summarization but require careful prompt engineering and data structuring. Visualization matters — charts make insights far more intuitive. Holistic wellness tools are powerful when they combine multiple life domains. Hackathon speed teaches the value of scoping features smartly.

What's next for Bestie

Integrate with real APIs (Fitbit, Plaid, Apple Health, etc.). Expand AI reasoning to handle long-term planning (multi-month budgeting & health progress). Strengthen privacy & security to ensure sensitive financial and health data is safe. Deploy to cloud (Render/Heroku + Vercel) for live testing.