DevLift

logo

Inspiration

AI tools like Cursor, Copilot, and ChatGPT make coding much faster. But many developers have started to feel like they are not “real developers” anymore. It is easy to paste large blocks of code without really understanding them. That takes away learning, confidence, and most importantly, ownership. When errors pile up without support or encouragement, frustration grows and burnout follows. We wanted to build something that restores a sense of ownership and makes coding feel rewarding again.

What it does

DevLift is a coding companion built to give developers back ownership of their work.

Explain-in-Your-Words: When a large block of AI-generated code is pasted, DevLift detects it and asks the developer to summarize it in one or two sentences using speech-to-text. This encourages reading through and understanding code instead of just accepting it.

Emotion-Aware Coding: Every few seconds, DevLift captures a frame from the webcam and runs it through an ensemble of five Roboflow models. A voting system selects the most reliable result, classifying the developer’s state as focused, frustrated, or happy (because the code runs!). Based on that, the companion responds with encouragement or celebration.

Error to Progress Feedback: DevLift monitors compilation and runtime errors. When one is fixed, it celebrates the achievement, reinforcing that the developer solved it themselves.

Session and Wellness Tracking: DevLift tracks coding streaks, sends hydration reminders, and provides a session summary showing time spent coding, errors fixed, and mood over time.

How we built it

We developed a VS Code extension for integration and UI.

We built a Roboflow-powered emotion recognition. Instead of relying on one model, we combined five public Roboflow emotion classifiers into a system that evaluates each frame in parallel. Each model outputs its prediction, and our voting layer compares the results to produce a consensus emotion. This ensemble-voting method filters out noise, smooths inconsistencies across models, and makes real-time classification far more robust. The result is a system that can reliably tell if the developer is focused, frustrated, or happy, and respond with encouragement or celebration at the right moment.

We integrated with the VS Code diagnostics API to monitor and celebrate error fixes.

We added timers and reminders for hydration and session tracking.

Challenges we ran into

Making emotion detection reliable. A single model was inconsistent across lighting and angles, so we engineered a five-model ensemble with a majority voting system to achieve consistent results.

Tracking the AI usage of the user was impossible due to the manner in which copilot and other AI's are integrated into IDE's. To overcome this we monitored large code inputs that most likely come from AI models or online forums. This ended up being a better solution as it covered more use cases in a simpler way.

Accomplishments that we're proud of

Designing a companion that focuses on confidence and accountability, not just productivity.

Engineering a robust emotion recognition system that combines five Roboflow models with voting to stabilize predictions.

Creating a tool not to combat, but to ensure that developers are able to keep up with the rise and speed of vibe-coding, and not get lost in the process.

What we learned

How to extend VS Code with diagnostics, reminders, and interactive features.

How to use Roboflow models to monitor user emotion.

Proper use of Git/Github in branching and pull requests.

That developer experience is about more than efficiency. It is about confidence, motivation, and ownership.

What's next for DevLift

We plan to expand DevLift into a collaborative platform where developers can connect and code together. Users will be able to share coding sessions, celebrate each other’s progress, and give real-time feedback. By combining emotion-aware guidance with social collaboration, DevLift aims to make coding not only more confident and rewarding but also more social and engaging. This next step will turn individual ownership into a shared, motivating experience.