Cypher
Cypher – One line ahead of every error.
The problem Cypher solves
Programming isn’t just about typing code — it’s also about focus, clarity, and emotional state. But coding can often feel lonely, stressful, and tiring. Many developers and students end up working long hours, often late into the night, feeling stuck or mentally drained. During those moments, what’s missing isn’t just technical help — it’s support.
Tools like GitHub Copilot help with writing code, but they don’t care how you're doing. They don’t know if you're tired, distracted, or frustrated. They respond to code — not to people. That’s where Cypher comes in.
Cypher is your real-time coding companion — not just smart, but emotionally aware and always present. It’s built to help you not just write better code, but to feel better while doing it.
Here’s what Cypher brings to your workspace:
Distraction Management
If Cypher notices you're picking up your phone or getting off-track, it gives a gentle reminder to stay focused. These nudges help keep you in flow and reduce wasted time.
Emotional Awareness
Using your webcam, Cypher can detect when you’re tired, frustrated, or down. If it sees something’s off, it responds with empathy — suggesting a break, offering encouragement, or just checking in.
Live Coding Help
Cypher can read your code and assist like a real teammate. Whether it’s fixing bugs, explaining errors, or giving feedback, it stays one line ahead of every issue you face.
Mentorship & Practice
Want to prep for interviews or improve your skills? Cypher can act like a technical mentor — asking questions, guiding your logic, or challenging you with real-world problems.
Feels Like a Real Presence
Cypher isn’t just text on screen. It shows up as an avatar in your browser, making the experience feel interactive and human — not robotic.
In short, Cypher is the coding buddy you wish you had — smart, kind, and right there when you need help the most.
Challenges we ran into
Model Hunting for Emotion & Object Detection
One of the biggest early challenges was finding the right model for emotion and object detection through the webcam. We experimented with multiple open-source models, local inference setups, and cloud APIs, but most didn’t give us reliable or context-aware results. After testing several options, we ended up using Gemini’s AI, which provided the most accurate and adaptable performance for our specific needs.
Webcam Feed to Backend Pipeline
Getting a smooth and efficient stream of webcam frames to the Flask backend — while ensuring minimal delay and acceptable frame quality — was tougher than expected. We had to carefully manage frame compression, threading, and performance tuning to avoid lags or model misfires.
Maintaining Real-Time Responsiveness
Since our assistant is supposed to react in real time, maintaining low-latency interactions between the webcam input, Gemini’s AI processing, and the UI feedback loop required significant backend optimization and async handling.
First-Time Hologram Setup
Setting up the hologram was one of the most exciting — and challenging — parts of the project. It was our first time building a working physical projection system out of theory. Aligning the projector, calibrating reflections on the glass pane, and making the avatar appear smooth and natural took multiple iterations. But in the end, we pulled it off, and the final result was definitely worth it.
Limited Time & Human Bottlenecks
With so many components working together — AI, backend, webcam integration, hologram display, and UI — we often found ourselves pressed for time when debugging new issues. And just when we were in full sprint, our team leader had a sudden bout of stomach issues and was, unfortunately, stuck in the bathroom for quite a while. That left us short-staffed during a critical stretch — but hey, we survived it, coded through it, and laughed through it too.
Tracks Applied (1)
Best use of Gemini API
Major League Hacking