Crowdgage

Inspiration

We noticed that with AI automating more of the technical work, people are spending less time sharpening essential soft skills, especially presentation and communication. Whether it's pitching a startup, presenting in class, or speaking at a meeting, being able to connect with an audience is still incredibly important. Yet, most people have no real way to measure how well they're doing. That got us thinking: What if there were a tool that gave you live, honest feedback, the kind your audience is thinking but not saying? That’s how Crowdgage was born.

What it does

Crowdgage is a device that evaluates key aspects of a presentation, including posture, body language, eye contact, voice clarity, emotional tone, confidence, and words per minute. It then translates that data into a simulated real-time audience displayed on an LED matrix grid. Green boxes represent engaged and interested audience members, while gray boxes indicate those who are disengaged. This live feedback allows presenters to adjust their performance on the spot, helping them improve their delivery and connect more effectively with their audience.

How we built it

We built our device using the ESP32 microcontroller and we built our AI models for presentation analysis using mediapipe, deepface, whisper AI, pytorch, and a fine-tuned ollama model for audience reaction simulation.

Challenges we ran into

One major challenge was ensuring real-time processing. We had to fine-tune our code to eliminate delays between the presentation input and the feedback output. Another challenge involved accounting for differences in speech patterns and accents, which required us to calibrate the system for more consistent results. Posture detection was also tricky due to hardware limitations, and we had to work hard to find a balance between providing useful insights and keeping the interface simple and user-friendly.

Accomplishments that we're proud of

We are proud that Crowdgage can accurately visualize audience engagement in real time using live presentation data. We successfully built a working device that delivers consistent and meaningful feedback. Beyond the technical success, we’re most proud of the impact the device can have. It’s not just a tool, but a coach that helps people become more confident and effective communicators. We also managed to keep the system intuitive and accessible, making it usable for students, professionals, and educators alike.

What we learned

Through this project, we learned how to apply machine learning and computer vision techniques to real-world human feedback scenarios. We discovered the importance of designing around the user’s needs, not just the technology. We gained experience in balancing data analysis with user simplicity and learned that even soft skills like public speaking can be improved with the right kind of real-time support. Most importantly, we saw how teamwork and iteration are key to turning an idea into a functioning and impactful product.

What's next for Crowdgage

As Crowdgage continues to evolve, we are focused on making it an even more powerful tool for developing real-world communication skills. One of our next goals is to integrate AI-driven coaching that can break down presentation performance and offer personalized suggestions, such as when to adjust pacing, how to improve tone variation, or where posture shifts affect engagement. We are also exploring the addition of interactive practice modes with dynamic audience simulations. These would allow users to rehearse in different virtual environments and receive adaptive feedback based on delivery style and content. To make Crowdgage more accessible, we plan to refine the hardware for better portability and usability in classrooms, offices, and personal practice setups. Our long-term vision is to turn Crowdgage into a go-to platform for presentation training, something that not only tracks how you are doing but helps you improve every time you use it.