Pro Se Pro: Empowering Plaintiffs for Small Claims

In small claims, 99% of litigants represent themselves, but many face a 80-90% failure rate against lawyers. Pro Se Pro uses Gemini to bridge this gap, turning evidence into trial-ready confidence.

Comment

Inspiration

In small claims, 99% of litigants go it alone, yet many face a staggering 80–90% failure rate when facing represented opponents. These individuals aren't choosing to be "Pro Se" (representing oneself) for fun; they are often low-income families or small business owners who cannot afford $300–$500/hour attorney fees.

We were additionally inspired by the "Justice Gap" - the distance between having the truth on your side and having the skill to present it. We also wanted to address "Rational Apathy": the tendency for people to abandon valid claims because the amount (e.g., $1000) doesn't seem to justify the "headache" of a lawsuit. We built Pro Se Pro to ensure that justice isn't reserved for those with deep pockets and to encourage people to stand up for their rights, no matter how small the claim.

What it does

Pro Se Pro is an AI-powered "Trial-Readiness Simulator" designed for the millions who navigate the court every year. Features at the MVP phase:

  • Case Intake: The journey begins with a multimodal intake engine. Users upload their court summons or initial complaints. The vision/multimodal feature of Gemini 3 parses these documents to automatically form structured output of the case information, jurisdiction, the legal cause of action, and the specific dates.
  • Evidence Engine: Uses Gemini 3 to recommend evidence to connect, and its vision capabilities** to parse physical evidence (receipts, photos, text messages). It flags Hearsay or Foundation issues and provides instant feedback to help users make their proof admissible in court.
  • Moot Court Simulator: A low-latency rehearsal space. Users face an AI Judge and Defendant that throw real-time objections, helping them practice their testimony, refine their narrative, and stay calm under pressure.

How we built it

Developed using Claude Code and Gemini 3 API. We leveraged Gemini 3 Flash’s low latency to create a "live" courtroom feel, allowing the AI Judge to interrupt and sustain objections, and AI defendant to interact with the plaintiff in real-time.

Challenges we ran into

  • The Simulator Loop: To mimic a real courtroom, we studied trial footage and consulted legal experts. We engineered a multi-agent system (Judge Agent and Defendant Agent) to ensure the simulation felt adversarial, reactive, and logically consistent.
  • The Knowledge Gap: As non-lawyers, we needed to ingest vast procedural rules without hallucinations. We used NotebookLM to synthesize complex YouTube trial walkthroughs and legal PDFs into structured insights. We then used Gemini’s 128k context window to "ground" our agents in these specific court manuals and county-level guides.

Accomplishments that we're proud of

We started this with little-to-no code background. This project is the result of "vibe coding" - using AI to describe our vision, logic, and empathy for the user, while letting the tools handle the technical syntax. Seeing a functional, high-stakes legal tool emerge from just a series of natural language conversations was our biggest "win."

What we learned

  • Prompting is the New Coding: We learned that being able to describe a complex logical flow (like the rules of evidence) is just as critical as knowing where the semicolons go.
  • Multimodal is Essential for Justice: We realized that legal tech isn't just about text; it’s about "seeing" a physical receipt or a photo of a broken sink and understanding its legal value as evidence.
  • The Power of Small Wins: Building this taught us that "good enough" AI tools can solve massive real-world problems today, even before they are "perfect."

What's next for Pro Se Pro: Empowering Plaintiffs for Small Claims

We’re just getting started on the road to democratizing justice:

  • Immersive Simulation: We plan to make the court simulator even more realistic by adding Voice-to-Voice interactions and Emotion Scoring to help users manage their tone and anxiety.
  • Expanding the Legal Brain: Currently, our knowledge base focuses on Illinois Small Claims. We plan to scale globally by ingesting hyper-local county rules, comprehensive statutes, and enriched rules depending on the **claim type" to provide "pro tips" tailored exactly to the courthouse where they will be appearing.
  • Real-World Feedback Loop: With some more UIUX polish-up and AI performance refinement, we aim at launching a closed beta for actual pro se litigants. By tracking real-world case outcomes, we will create a structured feedback loop to iteratively refine the AI’s strategic coaching.

Built With

Share this project:

Updates