Terminate Mate: Hackathon Project
A 24-hour innovation sprint to make legal knowledge accessible through AI
🚀 Original Hackathon Presentation & Demo
Terminate Mate was created for the Large Language Models (LLMs) x Law Hackathon #2 @Stanford University Law School, where teams were challenged to build solutions that make legal knowledge more accessible using AI technology. Our project explored how fine-tuned language models could help navigate complex employment termination laws in California.
Team Contributions:
- Henry Robinson: Built interactive prototype UI, scraped training data, and designed user flows
- Matt Helmer: Implemented all LLM fine-tuning and model training architecture
People admit to skimming through legal terms without fully reading them
Low-income Americans receive inadequate or no legal help for civil legal problems
Typical hourly rate for legal consultations, making legal advice inaccessible to many
The Problem
Note: This page presents the original hackathon pitch deck and prototype created during a 24-hour innovation sprint. The presentation style and content reflect the fast-paced, experimental nature of a hackathon environment rather than a formal case study.
One of the most frustrating aspects of dealing with legal matters is finding clear answers to seemingly simple questions. The law is inherently complex, and lawyers face significant downsides for giving wrong advice. Yet everyday individuals often find themselves bewildered due to the law's opaqueness.
For many people facing employment termination, understanding their legal rights can be overwhelming. Legal language is often deliberately complex, contradicting common sense expectations and creating barriers to justice for those who need it most.
The high costs of legal services (typically $200-500 per hour) mean that many people cannot afford to consult an attorney, even when they have legitimate legal concerns. This creates a justice gap where legal protections exist but remain inaccessible to the average person.
The Solution
Terminate Mate is an experiment to determine whether a fine-tuned large language model (LLM) can provide straightforward answers to common legal questions, specifically in employment termination law in California.
The system uses a specialized AI that has been trained on California employment law, legal precedents, and real-world cases to provide plain-language guidance that's accessible to everyone. By translating complex legal concepts into 9th-grade reading level language and presenting information in a conversational format, Terminate Mate makes legal knowledge more accessible.
Users can describe their situation in their own words, and the AI analyzes the details, identifies relevant legal concerns, and provides information about applicable laws and potential next steps—all while clearly indicating when a situation warrants consultation with a legal professional.
How It Works: The Jazz Musician Analogy
An LLM is like a jazz musician who has studied thousands of songs and musical patterns. When asked to play a new piece:
- They don't simply regurgitate memorized songs
- They understand the statistical patterns and relationships between notes
- They predict what notes would likely come next given the current musical context
- They generate new compositions that follow learned patterns but aren't copies
- Their output is influenced by both extensive training and the immediate context
Just as a jazz musician predicts the most harmonious next note based on musical patterns they've internalized, an LLM predicts the most likely next token based on the patterns in its training data. What makes Terminate Mate unique is that we've specialized this "musician" to perform specifically with California employment termination law.
Interactive Legal AI Solution
Explore this AI-powered legal assistant that I created myself using Claude and Next.js. This interactive prototype helps users understand their legal rights around employment termination in California through plain language conversation and real-world scenarios.
The Process
Research & Training
- Question Collection
Gathered the most viewed questions on termination from online platforms like Quora and Reddit
- Fine-tuning
Used ChatGPT 3.5 and fine-tuned it with legal precedents from the California Supreme Court
- Vector Database Creation
Combined answers to AI-generated questions with complete text of relevant California legal statutes
Design & Implementation
- Conversation Design
Created natural conversation flows to gather necessary information while maintaining user comfort
- Response Formatting
Instructed the AI to respond using 9th-grade-level language and reference relevant statutes
- UI/UX Design
Designed a simple, conversational interface with clear explanation of legal concepts
Business Model
Hybrid Revenue Model
For Users:
- Free basic legal assessment and information
- Premium subscription ($9.99/month) for unlimited consultations
- No obligation attorney referrals when needed
For Attorneys:
- Pay per qualified lead ($75-150)
- Enhanced visibility and profile features
- Access to anonymized case data and trends
Projected Annual Revenue
Next Steps
While the initial prototype focuses on employment termination law in California, the methodology can be expanded to cover other legal domains and jurisdictions:
- Expand Legal Precedents: Incorporate a broader range of legal precedents and case law
- Geographic Expansion: Adapt the methodology to cover all 50 states
- Diversify Legal Domains: Expand to landlord-tenant law, divorce law, peer-to-peer contract law, and small-claims law
- Enhance AI Capabilities: Improve the AI's ability to handle more complex legal questions and edge cases
- Develop Partnerships: Work with legal aid organizations and bar associations to validate and improve the service
By focusing on accessibility and plain language explanations, Terminate Mate represents a step toward democratizing legal knowledge and helping people understand their rights and options, without replacing the need for professional legal counsel when appropriate.