Inspiration - Freelancers lose time on invoices, payment tracking, receipts, and chasing overdue payments. I wanted an agent that does the work instead of just giving advice.

What it does - FreelanceFlow is an AI financial operations agent for freelancers. Users can simply say - “Rahul paid me ₹15,000 today.” The agent finds the client and invoice, records the payment, updates the balance, and generates a receipt. It also handles invoices, overdue payments, reminders, reports, and client balances.

How I built it - Built with Strands Agents SDK + Amazon Bedrock, with a FastAPI backend, Next.js dashboard and SQLite ledger. The agent uses 21 deterministic tools. The LLM handles reasoning and tool selection, while code handles money calculations, validation, database changes, and document generation.

Challenges I ran into - The biggest challenge was making an AI agent trustworthy with financial data. I had to prevent hallucinated clients, incorrect amounts, duplicate payments and inconsistent balances. I solved this with deterministic tools, validation, transaction-safe operations, and ledger-derived balances.

Accomplishments I'm proud of - 21 Strands tools 400+ automated tests Real invoice, payment, receipt, and reminder workflows Deterministic money handling Ambiguity and duplicate detection AgentCore-ready architecture End-to-end natural-language financial operations

What I learnt - I learned that effective agents aren't just about the model. The tools, data model, validation, and guardrails determine how safely an agent can act.

What's next for FreelanceFlow - I am taking FreelanceFlow toward production with Amazon Bedrock AgentCore, durable cloud storage, automated reminders, recurring invoices, email/calendar integrations and support for more small businesses.

My goal: tell FreelanceFlow what happened, and let it handle the rest.

Built With

Share this project:

Updates

Submission history