The Spark: Why MLT?
We were driven by a singular, urgent question: How can technology protect people when they are most vulnerable? We’ve all seen the headlines—tense, life-altering encounters between citizens and authority. While traditional safety apps simply ping the police, we realized that in the heat of the moment, what people really need is a clear head and a steady voice. We didn’t just want to build an app; we wanted to build a digital guardian that uses the power of AI to de-escalate situations and uphold human rights.
Introducing MLT: Intelligence for Human Safety
MLT isn't just a recording tool—it’s an intelligent safety companion. Imagine being pulled over and feeling that immediate spike of anxiety. MLT steps in to provide real-time, legal, and safety guidance. It analyzes the situation and whispers vital advice directly into your ear—reminding you of your rights and encouraging calm compliance—all synthesized into a seamless audio experience.
Our sophisticated safety pipeline:
- Constant Vigilance: The app maintains a secure, rolling one-minute audio buffer, ensuring no critical context is ever lost.
- Hardware Empowerment: With a simple press of the volume button, the user triggers a high-stakes analysis.
- Intelligent Processing: Audio is transcribed and speaker-diarized via ElevenLabs, creating a clear map of the interaction.
- Dual-Agent Logic: We deployed two custom-instructed Gemini Agents. The first acts as a scout, identifying if a situation is escalating into danger. If the "red alert" is triggered, the second agent instantly generates life-saving advice and legal reminders tailored to the specific conversation.
The Engine Under the Hood
To match the urgency of our mission, we built a backend that is lightning-fast and rock-solid. We leveraged Elysia.js on the Bun runtime for peak performance, containerized the entire architecture with Docker, and deployed it via DigitalOcean.
The frontend was crafted in Flutter, ensuring a seamless, life-saving interface is available to both iOS and Android users. We didn't just write code; we engineered a bridge between hardware and human safety.
Triumph Through Turmoil
The road to MLT wasn't easy. We faced intense debates, stripping away distractions to protect the "core soul" of the app. We fought against the clock and the limitations of remote collaboration, splitting into two focused strike teams (Frontend and Backend) to maintain momentum. We had to be ruthless—scrapping good ideas to ensure the Minimum Viable Product was more than just functional; it had to be flawless. We didn't just finish a project; we delivered a proof of concept for a safer world.
The Future of "Vibe-Coding"
Our development process was a testament to the future of software. We used Google’s Antigravity to push the boundaries of AI-assisted creation. We "vibe-coded" the frontend, prompting an entire interface into existence through pure vision. On the backend, we used AI as a high-speed co-pilot, allowing our developers to focus on high-level logic while the AI handled the heavy lifting.
MLT is more than an entry in a hackathon—it’s a vision for a world where technology works to keep us calm, kept, and safe.
Log in or sign up for Devpost to join the conversation.