Geo-Compliance Classifier

Inspiration

We were inspired by the critical need to transform regulatory compliance from a reactive, manual burden into a proactive, automated safeguard. We wanted to empower product teams to innovate confidently by identifying geo-specific legal risks before a feature even launches.

What it does

The Geo-Compliance Classifier is an internal tool that uses AI to automatically screen product feature descriptions. It flags those requiring geo-specific compliance logic, generates a detailed audit trail with reasoning, and streamlines the manual review process with integrated reporting.

How we built it

We built a full-stack application using Streamlit for the frontend. The backend leverages Google's Gemini API for LLM analysis, FAISS and sentence-transformers for semantic similarity checks, and Web3.py to log immutable records on the Ethereum blockchain. We integrated Slack for team notifications using the Bolt framework.

Challenges we ran into

Our main challenges were refining the LLM prompts to prevent hallucination and ensure consistent, structured outputs, managing API rate limits and costs during batch processing, and designing a system to log data to the blockchain in a cost-effective and efficient manner.

Accomplishments that we're proud of

We're proud of creating a functional prototype that successfully automates a complex legal analysis task. We built a seamless pipeline from user upload to AI classification, immutable blockchain logging, and team collaboration via Slack, all wrapped in an intuitive user interface.

What we learned

We gained deep, practical experience in prompt engineering, configuring and throttling API calls, and integrating disparate technologies (AI, blockchain, SaaS platforms) into a single, cohesive application. We learned to treat the LLM not as an oracle, but as a powerful tool that requires careful guidance and validation.

What's next for Geo-Compliance Classifier

Next steps include expanding the knowledge base of regional regulations, implementing a feedback loop to continuously improve the AI model, adding more advanced analytics to the dashboard, and exploring fine-tuning a custom model for even greater accuracy and cost-efficiency.