Problem
Have you ever frozen in front of a room full of people, with your mind going completely blank? I found this problem when I started high school. Public speaking was my absolute worst nightmare. To survive, I started memorizing scripts word-for-word. But every single time I stood up, I forgot everything, and my improvisation skills were even worse. Through a lot of painful trial and error, I finally realized something important: a great presentation isn't about having gorgeous slides. It’s about you. Over 70% of people struggle with public speaking anxiety. We spend hours perfecting our visuals and learning our scripts, but we completely ignore what actually matters: our voice, our pace, and our body language. When you want to practice, you either talk to a mirror, or a friend tells you, "Yeah, it was fine." But that doesn’t help. Nobody tells you in real-time that you're speaking too fast, stumbling over your words, or standing with bad posture.
Solution
That’s why I built PitchIt—your personal presentation coach, right in your pocket. You choose what you want to practice, whether it's an investor pitch or a casual talk. The app gives you 5 minutes to read a short article, and then you start speaking while your device watches and listens. It instantly tracks your pacing and counts your filler words. Thanks to AI and your webcam, it also analyzes your body language and posture. Within seconds, you get honest, easy-to-understand feedback just like you're working with an expert mentor.
How I built it
The best part? I didn't spend months writing thousands of lines of code by hand. I built the whole thing using AI prompting, turning my idea into a working MVP super fast. To make sure the AI doesn't hallucinate or make stuff up, I split the logic: hard data like Words Per Minute (WPM) are calculated right on the frontend using code. The LLM (running on LLaMA via Groq for instant speed) is used strictly for the qualitative feedback and context.
Challenges
One of the biggest challenges was making sure the AI didn't just guess the feedback. The biggest headache was figuring out how to stop the AI from making things up or guessing random numbers. I had to learn how to separate the actual math (like counting words per minute in the code) from letting the AI just handle the feedback text. It was a huge challenge to me.
Next steps
I’ve lived through this fear myself. Today, the product is fully functional, and I am ready to launch it for testing. Speaking in public shouldn't feel like torture. With PitchIt, it becomes your strongest superpower.
Log in or sign up for Devpost to join the conversation.