MiddleMe

Inspiration

In a world increasingly opposed by bottlenecks, we aren't just losing arguments; we’re losing relationships. We noticed that most communication tools are designed for speed or broadcast, but none are designed for reuniting relationships. We wanted to build a "digital neutral ground" that moves users past the "who is right" mentality toward shared understanding.

What it does

Middle Me is an AI-powered mediation platform that helps people resolve disagreements through a structured emotional journey. Users begin with perspective onboarding to define their core values, or they can "Skip to the Middle" to address an immediate conflict. The core "Meet in the Middle" feature allows users to explain their side or upload conversation media, which the AI then de-escalates by identifying shared values and asking reflective questions to increase the user's emotional score. For professional environments, a dedicated Business Mode functions as an Employee Assistance Program (EAP), using RAG over company policies to provide mediation that is both empathetic and policy-aware.

How we built it

We optimized our technical stack for demo clarity and high reliability within the hackathon timeframe. The frontend is built with Next.js App Router, Tailwind CSS, and shadcn/ui to ensure a polished, low-cognitive-load user experience. For the backend, we utilized Supabase for authentication and a Postgres database with pgvector to power our RAG pipeline. The intelligence layer is driven by Gemini AI integration and Edge Functions, allowing us to process multi-modal inputs like text, images, and video into neutral, de-escalated insights.

Challenges we ran into

One of our primary hurdles was fine-tuning the AI mediation pipeline to ensure it remained truly neutral and empathetic without appearing condescending. We also had to manage the complexity of converting various media formats into clean text using the Google API while maintaining the original context of the dispute. Furthermore, balancing a comprehensive onboarding flow with the need for a fast-paced hackathon demo required careful "Scope Control," forcing us to decide which features to polish and which to simplify for maximum impact.

Accomplishments that we're proud of

We are particularly proud of our "AI Experience" logic, which successfully moves beyond simple summarization to identify misunderstandings and bridge statements. Successfully integrating the Business Mode EAP demonstrated that our mediation UX could scale from personal relationships to professional corporate structures. Finally, we achieved a high-fidelity "Emotional UX" that prioritizes the user's feelings and relationship health over technical complexity

What we learned

This project taught us that the true "pain point" in most conflicts isn't a lack of information, but a lack of facilitated perspective-taking. Technically, we learned how to architect an AI system that prioritizes "Emotional Scores" as a primary metric for success. We also discovered a plan focusing on avoiding overengineering and premature scaling is the most effective way to deliver a functional, high-impact demo under pressure.

What's next for MiddleMe

Looking ahead, we plan to implement a "For You" feed that offers personalized wellness and social connection advice based on a user's historical perspectives. We also aim to expand the Business Mode RAG capabilities to handle more complex, multi-layered corporate handbooks. Ultimately, we want to evolve Middle Me into a real-time assistance tool that can help de-escalate tension in live conversations as they happen.