AI Wealth Pro

An AI first personal finance app which allows you to make an LLM smarter by sending it a summary of your financial status and help give personalized advice

Comment

Inspiration

I recently moved from India to Ireland and discovered something surprising:
even the wealth manager bundled with my relocation package couldn’t answer my “simple” cross-border money questions as well as ChatGPT could—once I fed it the right prompts and data.

That experience sparked two insights:

  1. LLMs can beat human advisors — but only if they’re armed with accurate, granular financial data.
  2. Most people will never spend hours iterating on prompts or manually pulling CSVs.

Budget apps, static Excel sheets, even dedicated advisors still leave a gap between “raw numbers” and clear, timely guidance. AIWealth.me bridges that gap by syncing live bank data and letting an AI advisor surface actionable, context-aware recommendations instantly—zero CSVs, zero prompt engineering required.

What it does

AIWealth.me (working title: AIWealth Pro) is an AI powered financial intelligence platform:

  • Live Bank Sync – Plaid connects 11k+ institutions; balances and transactions flow in automatically.
  • AI Advisor – Our edge-function calls multiple LLM's (Perplexity for now) to turn raw data into plain-English insights, goal recommendations, and risk alerts. The prompt that we use is a fine-tuned variation of one of the most recommended prompt in personal finance and FIRE industry.
  • Portfolio Tracker – Track 401k, brokerage, and crypto in one place, with automatic allocation and gain/loss breakdowns.
  • Goal Engine – Create goals (emergency fund, vacation, down payment) and see real-time progress plus AI-suggested contribution schedules.
  • Subscription Scanner – Flags recurring charges.

How we built it

  • Frontend – React 18 + TypeScript, Tailwind CSS, Lucide icons.
  • Backend – Supabase Postgres with Row-Level Security and Edge Functions.
  • Bank data – Plaid Link for OAuth, Plaid webhooks into a Supabase function.
  • AI layer – A single Supabase Edge Function (ai-chat) that calls Perplexity now and is designed to swap in OpenAI/Anthropic later.

Challenges we ran into

  1. Webhook race conditions – Plaid sometimes delivered duplicate transactions; we solved this with a unique constraint + upsert logic.
  2. Per-user data isolation – RLS rules had to cover 8 tables; one mis-configured policy broke the entire dashboard. Took bolt a lot of time to get this done.
  3. Prompt design – Getting Victor Sterling to sound helpful but not “advisory” took multiple iterations and a lot of testing.

Accomplishments we're proud of

  • Live product - the website's live. You can use it, and it's pretty amazing that we've been able to get it with all the features.
  • demo: connect a sandbox bank, ask “How’s my spending this month?”, and Victor replies in <5 sec.
  • Zero PII in logs—tokens are encrypted at rest, rotated every 30 days.

What we learned

  • People do not want to connect their bank and ChatGPT, they are afraid because ChatGPT is too powerful.
  • Any individual bank is better suited to build what we are building, but on an average, an individual has three or more bank accounts.
  • Finance users care more about specific next steps than pretty charts.

What's next for AIWealth Pro

  1. Multiple AI providers – Add OpenAI + Anthropic fallback to reduce vendor risk.
  2. Automated budgeting – Let Victor auto-allocate income to envelopes.
  3. Mobile Apps – Push-notification alerts for large transactions or goal milestones.
  4. SOC 2 prep – Begin evidence collection for production launch.

Built With

Share this project:

Updates

Submission history