Ratio

Inspiration

Conflicts waste time, drain energy, and create tension in teams. We wanted to build Ratio, an AI-powered conflict resolver, to make disagreements less destructive and more constructive. The idea came from observing how much time workplaces lose on avoidable miscommunication, and how difficult it is for people to mediate without bias.

What it does

Ratio helps people resolve conflicts by acting as an AI-powered mediator. Collects Perspectives: Both sides share their viewpoints in a structured way. Analyzes Language: The AI identifies pain points, emotional triggers, and areas of agreement. Generates Mediation Strategies: Provides tailored suggestions such as compromises, reframing techniques, or step-by-step resolutions. Guides Conversation: Offers prompts to keep the discussion productive and neutral. Outcome Tracking: Summarizes resolutions so both parties have a clear record of agreements.

How we built it

We built Ratio with a focus on speed, neutrality, and usability: Backend: Used Python (FastAPI) to handle requests and serve the AI pipeline. AI Layer: Integrated open-source LLMs (via GPT OSS/Ollama) with custom prompts to analyze perspectives and generate mediation strategies. Frontend: Designed with React + Tailwind CSS for a minimal, distraction-free interface. Database: Leveraged Supabase for secure storage of user inputs, conversation logs, and outcomes. Architecture: Deployed as a modular system so the AI engine, database, and frontend can scale independently. Workflow: Iterated quickly using Agile sprints, frequent testing, and real-time feedback loops to refine mediation strategies.

Challenges we ran into

Bias in AI responses: Making sure the system didn’t “side” with one participant. Handling emotional language: Designing prompts that avoid escalating sensitive conversations. Time constraints: Building a functional MVP while testing multiple approaches. Technical hurdles: Integrating AI inference with real-time response requirements.

Accomplishments that we're proud of

Built a working MVP of an AI conflict-resolution tool in a short timeframe. Designed a system that analyzes perspectives without bias, keeping both sides engaged. Created a clean, distraction-free UI that makes a tense process feel approachable. Learned how to integrate open-source LLMs effectively for real-world use cases. Managed to turn an abstract idea—AI as a mediator—into a practical, testable solution.

What we learned

AI isn’t magic: It needs careful prompting and bias checks to handle sensitive conflicts. User experience matters: Even the smartest AI fails if the interface feels intimidating. Neutral tone is hard: We learned how to design language that feels balanced and fair. Time discipline: Hackathon deadlines forced us to prioritize core features over “nice-to-haves.” Collaboration skills: We improved at dividing tasks, syncing quickly, and resolving our own team disagreements (ironically).

What's next for Ratio

We plan to add real-time mediation, support for multi-party conflicts, and integrations with tools like Slack and Teams, while improving accuracy, privacy, and emotional intelligence.

Built With

• ai
• api
• gpt
• javascript
• node.js
• python
• react
• tailwind