Hey everyone,
I’m part of a team at “Modified by moderator”, and we’re currently working on a next-gen AI-powered accounting software development that leverages Gemini (via Vertex AI/Gemini API) to automate, simplify, and enhance financial operations for small and mid-sized businesses.
We’re aiming to embed smart features such as:
- Context-aware financial report generation
- Real-time natural language query support for ledgers, invoices, etc.
- Smart expense categorization using LLM understanding
- Conversational AI assistant for tax, compliance, and reconciliation queries
- AI-driven anomaly detection in transactions
Our tech stack:
- Frontend: Flutter
- Backend: Node.js + Firebase (Firestore/Auth)
- AI/ML: Google’s Gemini API (in active prototyping)
- Accounting Engine: Custom logic integrated with standard financial APIs
While the core functionality is coming together, we’re currently exploring and experimenting with Gemini for the intelligent layer — and running into a few challenges:
Designing effective prompts for financial domain-specific tasks
Managing Gemini’s streaming responses in chat-style financial assistants
Maintaining accuracy when parsing context-rich, ambiguous queries
Linking AI-generated output to real-time transactional and ledger data
Optimizing performance and cost when scaling across many users with dynamic queries
We’re building this from both an innovation and usability perspective, aiming to create a powerful tool that’s actually helpful in the day-to-day of accounting workflows.