Revellio

Inspiration

Managing money for international students is almost a second job. We face an "Anxiety Loop" by managing bills in USD, savings in BRL, and facing a constantly shifting FX market. Revellio was created to manage this global financial life, offloading the work and allowing for international students to relax.

What it does

Revellio is an agentic platform that predicts debts, identifies optimal conversion rates, and .

Visionary Accountant: Uses Optical Character Recognition to process invoices and lease agreements, extracting dates and amounts into a Liability Ledger.
FX Strategist: Monitors FX rates and uses Technical Analysis (RSI/MACD) to identify market behavior and optimal exchange opportunities.
Smart Router: Performs real-time simulations of different transfer options (Wire, Crebit, Remitly) to find the most efficient path accounting for transfer urgency, fees, and amount.
Trust Engine: Generates a "Proof of Reason" reasoning packet for every action, hashing it to a public ledger for an unchangeable audit trail.

How we built it

The system is built as a Stateful Multi-Agent Graph to allow specialized agents to share context.

Orchestration: LangGraph manages the agentic loops and states.
Intelligence: Gemini 3.1 for market prediction, financial data ingestion and Optical Character Recognition (OCR).
Blockchain: Stellar Network serves as our immutable "Digital Notary." We utilize a three-layer Trust Engine that generates a cryptographic fingerprint of every financial decision.

Challenges we ran into

Integrating real-time financial data with agentic reasoning required overcoming significant latency and consistency hurdles. We specifically struggled with ensuring the "Audit Hash" remained perfectly synchronized between our local Postgres database and the Stellar ledger. We solved this by implementing an optimization check that prevents ledger spam during stagnant market conditions while ensuring instant finality whenever a high-impact financial decision is executed.

Accomplishments that we're proud of

We improved the "Black Box" problem of AI, especially in such a sensitive topic as finance, by implementing a Proof of Reason audit trail. Users can verify exactly why an agent made a decision, such as seeing that the BRL was "Oversold" during a specific prediction.

What we learned

We learned the importance of and the steps towards implementing transparency in agentic systems. We also deepened our understanding of agent orchestration along with the intricacies of the FX market.

What's next for Revellio

Our main goals going forward are to offer further automation capabilities to the agents, allowing them to automate payments under the user's discretion and permission. Along with this push for further agency, we look forward to implementing increased reasoning transparency and financial safeguards.