Inspiration
The most unpredictable part of our lives are not the mechanical elements or codebases, but rather, it’s the people. Conversations are by far the hardest skill to practice, yet they determine not only our friendships, but also our relationships and careers. In fact, some research shows that 71% of Gen Z-ers struggle to make small-talk with co-workers. And many other studies show a ‘fear’ of talking to potential romantic partners. That inspired me to build the Life Coach: a tool designed to make conversations seems a little less intimidating and more approachable, especially for younger generations learning to communicate in different ways when so much is done on screen.
What it does
Life Coach is a conversational practice partner. Users can speak to it as if they’re preparing for a date, a job interview, or even just small talk. The app listens, responds with natural, adaptive dialogue, and provides real-time coaching, telling them how to improve and potential ways to keep the conversation going. It has safeguards so users can’t be explicit, teaching them the nuances of communication, and can be easily adjusted for the various scenarios that the user may come across.
How we built it
I built the app using Python with a Flask backend, running LLaMA as the generative AI model. For real-time communication, I integrated the sounddevice library to handle voice input and output, making the experience feel like a live conversation instead of a text exchange. Dependencies were managed in a .venv environment.
Challenges we ran into
One of the biggest challenges was ensuring smooth real-time audio without lag. Another was getting generative AI to stay conversational without becoming too verbose or robotic—which it has a tendency to do. On top of that, deployment was trickier than expected because of library compatibility issues across different systems.
Accomplishments that we're proud of
I’m proud that the prototype doesn’t just “work” technically—it feels actually human. I got some of my friends to be better at phrasing something after a few minutes of practice, which was the best validation. More than building another AI app, I built something that could make a genuine difference in how people connect with each other, we can all use practice communicating verbally sometimes, especially as so much of our communication is via Direct Message or Messages.
What we learned
I learned that the hardest part of building AI apps isn’t the technical side, but about getting the technology to work with actually solving the problem. No matter what, the most complex variable in any system is still people, and building for that is certainly difficult.
What's next for Life Coach
Next, I want to expand the app with even more coaching modes, from dating to workplace communication. (Right now there are three.) Personalization is also a big priority of mine by letting the AI adapt to a user’s tone, comfort level, and growth over time. A mobile-first version would make it accessible in real-world contexts, like before a first date or a job interview. I also want to integrate measurable “confidence scores” based on communication research, so users can see tangible progress.
Log in or sign up for Devpost to join the conversation.