Inspiration The inspiration for Contract Co-Pilot comes from a recurring problem faced by thousands of students at Ontario Tech University and across the province every September. Many students and first-time renters are presented with complex, multi-page lease agreements filled with "legalese". Landlords frequently include illegal or unenforceable clauses—such as blanket "No Pets" policies or illegal damage deposits—betting on the fact that tenants do not know their rights under the Ontario Residential Tenancies Act (RTA). We wanted to build a "shield" for tenants that levels the playing field using AI.
What it does Contract Co-Pilot is an AI-powered legal auditor designed specifically for Ontario's jurisdiction.
Instant Risk Assessment: Users upload their lease PDF, and the system instantly scans it for "Red Flag" (illegal) and "Yellow Flag" (risky) clauses.
Jurisdiction-Aware Auditing: Unlike generic AI, it cross-references the text against the official Ontario RTA (2006) to verify legality.
Hybrid RAG Chat: Tenants can ask natural language questions like "Is this damage deposit legal?" and receive answers cited directly from provincial law.
Safety Guardrails: The system identifies dangerous advice (like withholding rent) and directs users to the correct legal forms, such as the LTB T6 Application.
How we built it We utilized a modern, scalable RAG (Retrieval Augmented Generation) architecture.
Backend: Built with Python and FastAPI for high-performance, asynchronous processing.
AI Engine: Leveraged Google Gemini 3 Flash for low-latency, intelligent analysis and synthesis.
Database: Used ChromaDB as a vector store to embed and retrieve sections of the Ontario RTA and the user's specific lease.
Frontend: A clean, responsive interface built with HTML5, CSS, and Vanilla JavaScript to ensure a frictionless user experience.
Challenges we ran into The biggest technical hurdle was Jurisdictional Accuracy. Generic LLMs often hallucinate by mixing American laws with Canadian ones. We solved this by implementing a RAG pipeline that "forces" the AI to retrieve and prioritize the local Ontario RTA statutes before generating a response. Another challenge was handling "Schedule A" or "Additional Terms" pages, where landlords often hide illegal hand-written or custom-typed rules. We refined our PDF parsing to ensure these custom sections were thoroughly audited.
Accomplishments that we're proud of We are particularly proud of our "Privacy by Design" approach. In a world of data harvesting, we chose to implement no login or signup requirement. Leases are processed in memory and instantly flushed; zero data retention means zero risk for vulnerable tenants. We also successfully built a tool that provides the specific LTB form names (like the T6) needed for resolution, moving beyond simple summaries to actionable help.
What we learned This project taught us the power of Contextual Grounding. We learned that for AI to be useful in high-stakes fields like law, it cannot simply be a "black box"—it must show its work by citing specific statutes. We also gained deep experience in building a full-stack AI application from scratch, from managing virtual environments and git history to optimizing prompt engineering for legal disclaimer generation.
What's next for Contract Copilot We want to expand the "Knowledge Base" to include Tenant Board (LTB) case law to predict the outcomes of disputes. Additionally, we plan to add support for Automated Form Filling, where the AI can help a tenant pre-populate an LTB application based on the illegal clauses it detected in the lease. Our goal is to make Contract Co-Pilot the go-to resource for every student renter in Ontario.
Log in or sign up for Devpost to join the conversation.