MentorMoney – AI Financial Wellness Coach (FinTech track)

MentorMoney turns banking data into Gemini-powered coaching so Maya and other young pros stay on budget, weigh big decisions, and learn just-in-time lessons that build lasting confidence.

Comment 1

Inspiration

  • Team members saw friends stress-scroll through banking apps without knowing what to do next; we wanted an AI mentor that explains numbers in plain English and addresses bias directly.
  • Hackathon theme focused on FinTech & Financial Literacy, which aligned with our goal to meet Gen Z where they are and reduce anxiety around money decisions.

What it does

  • Ingests persona financial snapshots and produces Gemini daily digests with actionable tips and recommended lessons.
  • Offers a scenario sandbox (rent vs. buy today, more calculators coming) with Gemini narratives that translate numbers into trade-offs.
  • Generates custom micro-lessons + quizzes on the fly, stores them in the curriculum, and surfaces them in the Learning Hub.
  • Provides live financial controls so updating income/expenses instantly re-runs insights and shows new surplus metrics.

How we built it

  • Backend: FastAPI + Python handles persona storage, calculators, and Gemini prompt orchestration (Gemini 2.5 Flash for insights, custom prompt for lesson generation).
  • Frontend: React + Vite SPA with Zustand for state, Tailwind styles; dashboard consumes APIs, renders Markdown (React Markdown + custom normalizer).
  • Added resilience layers: fallback mocks, text cleaners, and Markdown normalization to tame Gemini output.
  • Tests: Pytest for backend flows, ESLint/TypeScript checks for UI.

Challenges we ran into

  • Gemini sometimes returned partial or malformed JSON; we had to sanitize outputs, enforce schema defaults, and synthesize quizzes when missing.
  • Markdown formatting arrived with broken headings (“Overvie\nw”); required regex normalization to fix sections without losing content.
  • Keeping persona recommendations in sync after generating new lessons, which was solved with store updates and caching.

Accomplishments that we're proud of

  • End-to-end flow: enter direction → Gemini generates on-topic lesson + quiz → instantly appears in Learning Hub.
  • Daily digest now surfaces both actionable next steps and lesson recommendations, making insights immediately usable.
  • Scenario narratives are cleanly formatted Markdown with consistent headings and bullets after heavy post-processing.
  • Runbook and pitch assets document the full journey for judges, including validation and fairness logs.

What we learned

  • Structured prompts + deterministic post-processing are essential when relying on LLM output for production-like flows.
  • Providing Gemini with explicit resolved context (persona summary, track hints) drastically improves topical accuracy.
  • Small UX touches, spinner states, financial controls, and Markdown cleanup make AI insights feel trustworthy.
  • Continuous testing (pytest, lint, build) prevents regressions as we iterate quickly.

What's next for MentorMoney – AI Financial Wellness Coach

  • Add more calculators (debt snowball, emergency runway) and expose them via the scenario sandbox.
  • Integrate real banking APIs (Plaid/Teller) for live data and expand persona creation (custom financial inputs + goal tagging).
  • Enhance fairness auditing and multilingual support; deliver bilingual insights consistently.
  • Package the web client into a mobile-friendly PWA and prepare for pilot deployments with university/employer partners.

Built With

Share this project:

Updates