Inspiration
Managing personal finances is often fragmented across bank statements, budgeting apps, news platforms, spreadsheets, and financial planning tools. We built WealthPilot to bring those experiences together into one AI-powered assistant that can help users understand their finances through natural conversation, voice commands, visual insights, and goal tracking.
Our goal was to make financial information easier to understand without requiring users to manually organize every transaction or navigate complicated dashboards.
What it does
WealthPilot is an AI-powered personal finance assistant that helps users:
- Import PDF bank statements and automatically extract transactions
- Detect duplicate and overlapping transactions during statement imports
- Categorize spending and generate dynamic financial charts
- Ask financial questions through a conversational AI assistant
- Use voice commands and the “Hey Assistant” wake phrase
- Receive spoken AI responses using text-to-speech
- Create and update savings goals through the AI assistant with user approval
- Track spending, savings, budgets, and financial goals
- View financial and industry news with voice navigation
- Secure user accounts and financial data using Supabase authentication and Row Level Security
- Use a dark, mobile-first interface designed to reduce financial information overload
How we built it
WealthPilot was built as a Flutter mobile application with a Node.js/Express backend.
The mobile application communicates with our backend API for AI-powered financial analysis. OpenAI is used for conversational intelligence and financial reasoning, while custom backend logic handles financial tools, statement parsing, chart planning, goals, and news.
Supabase provides authentication, user profiles, transaction storage, statement import history, upcoming bills, financial goals, and Row Level Security so each user can only access their own data.
We also built custom logic for:
- PDF transaction extraction
- Duplicate transaction detection
- AI-generated visualization planning
- Goal proposal and approval flows
- Voice command routing
- Wake-phrase detection
- Text-to-speech
- Financial news aggregation
The backend is deployed publicly on Render, and the Android application is distributed as a signed APK through GitHub Releases.
Challenges we ran into
One of our biggest challenges was making multiple AI and voice systems work reliably together.
Voice recognition sometimes captured commands correctly but failed to submit them to the AI. The wake phrase also had to continue listening while users moved between different screens.
News narration introduced another challenge: after opening one story, the speech recognizer needed to restart so users could immediately say commands such as “open news 2.” We also had to ensure narration stopped when users navigated back.
Dynamic chart generation required careful coordination between AI responses and visualization planning so financial questions produced appropriate graphs instead of irrelevant visualizations.
Statement imports also required robust duplicate detection. We had to distinguish genuinely new transactions from overlapping transactions across multiple uploaded statements.
Goals introduced additional complexity because the AI should never silently modify financial goals. We designed an approval flow so the assistant proposes a change and the user explicitly confirms it.
Finally, moving from local development to a downloadable APK required configuring Supabase, production API environment variables, release signing, and deploying the backend to a public HTTPS endpoint.
Accomplishments that we're proud of
We are proud that WealthPilot became much more than a chatbot.
The final application combines conversational AI, voice interaction, financial charts, PDF statement processing, duplicate detection, goals, financial news, authentication, and persistent user data in one mobile experience.
Some accomplishments we are especially proud of include:
- A working voice-enabled AI financial assistant
- “Hey Assistant” wake-phrase support across the main application
- Dynamic financial visualization generation
- Reliable statement overlap and duplicate detection
- AI-assisted goal creation and updates with human approval
- Voice-controlled financial news navigation
- Secure Supabase-backed user data
- A signed Android APK that can be downloaded and installed directly
What we learned
We learned that building a useful AI product requires much more than connecting an application to a language model.
The most important work was designing reliable systems around the AI: structured financial tools, approval flows, persistent data, visualization planning, speech recognition, navigation state, and error handling.
We also learned how important human control is when AI can modify financial information. Instead of allowing automatic changes, we designed WealthPilot so important actions such as goal updates require explicit user approval.
On the engineering side, we gained experience with Flutter, Node.js, Supabase, OpenAI APIs, speech recognition, text-to-speech, PDF processing, authentication, Row Level Security, Android release signing, and cloud deployment.
What's next for WealthPilot
Next, we want to expand WealthPilot into a more complete personal financial intelligence platform.
Future improvements include:
- Production subscription and premium-access support
- Better automatic transaction categorization
- More advanced forecasting and cash-flow analysis
- Personalized recurring-bill detection
- Improved financial goal recommendations
- Additional chart types and interactive visualizations
- More financial news sources
- Improved voice recognition and natural-language navigation
- iOS support
- Stronger privacy controls and financial-data integrations
Our long-term vision is for WealthPilot to become an intelligent financial co-pilot that helps users understand their money, plan confidently, and make more informed financial decisions.
Built With
- ai
- android
- api
- codex
- dart
- express.js
- flutter
- github
- javascript
- node.js
- openai
- postgresql
- processing
- python
- recognition
- render
- rest
- speech
- sql
- supabase
- text-to-speech

Log in or sign up for Devpost to join the conversation.