SimTeach | Devpost

Inspiration

As high school students, we’ve seen first-hand how much impact teachers have on our learning. The best teachers made classes engaging, clear, and motivating. But we’ve also experienced the other side—teachers who struggled with explaining concepts, managing a classroom, or keeping students engaged. Many of them were clearly passionate, but lacked practical training opportunities before stepping into the classroom. Unlike pilots or doctors, who can practice in simulators, teachers often have to learn by trial and error—with real students. That means we, as learners, sometimes pay the price for that lack of preparation.We were inspired to build SimTeach as a way to fill this gap: a safe, AI-powered teaching simulator where educators can practice, get feedback, and grow their skills before stepping in front of a class.

What it does

SimTeach is an AI-driven classroom simulator designed to give educators a safe, responsive space to practice and refine their teaching skills. Through an interactive video-call interface, teachers conduct lessons with a dynamic AI student programmed to emulate real high school behaviors—whether it’s asking thoughtful questions, expressing confusion, or making common mistakes in subjects like algebra, such as factoring or exponent rules.

Every session is transcribed in real time, and once the lesson concludes, SimTeach generates detailed, structured feedback across essential teaching competencies. This includes rapport-building, questioning techniques, adaptability in explaining concepts, and subject knowledge clarity. Teachers receive actionable insights highlighting their strengths and areas for growth, along with suggestions for what to focus on in future sessions.

All session data—transcripts, feedback reports, and performance metrics—are stored securely and accessible through a personalized dashboard. This allows educators to track their progress over time, identify patterns, and continuously improve with evidence-based guidance. By blending realistic AI interaction with in-depth analytics, SimTeach serves as a powerful professional development tool, helping teachers build confidence and competence before stepping into the classroom.

How we built it

SimTeach was built using a modern, full-stack framework to ensure scalability, real-time interaction, and powerful AI-driven feedback. The frontend was developed with Next.js, TypeScript, and Tailwind CSS, providing a responsive and intuitive user interface. For authentication, we integrated Auth0 to securely manage user access and sessions.

The core of our platform is powered by advanced AI models. We leveraged Gemini and ChatGPT to create realistic, dynamic student personas capable of natural, context-aware interactions during teaching sessions. To capture and analyze teaching sessions in detail, we used Tavus for video recording and behavioral analytics, enabling features like facial expression tracking and engagement metrics.

On the backend, we used MongoDB as our primary database to store user profiles, session transcripts, feedback reports, and long-term progress data. All session interactions—including live video and real-time feedback—are processed through a secure and efficient serverless architecture.

This integrated tech stack allows SimTeach to deliver a seamless, immersive practice teaching experience complemented by deep, actionable insights.

Challenges we ran into

We faced several significant technical hurdles while building SimTeach. The integration of Tavus proved particularly complex, as synchronizing real-time video streams with our AI feedback system required precise timing and extensive debugging to ensure seamless operation.

Processing and storing data also presented a major challenge. Converting voice data from teaching sessions into accurate transcripts, then parsing, analyzing, and storing that information into MongoDB in a structured and efficient way demanded careful pipeline design and error handling.

Additionally, blending multiple AI systems—including Gemini and ChatGPT—into a cohesive and responsive student persona required fine-tuning and prompt engineering to maintain consistent and realistic interactions. Ensuring the system remained scalable while managing real-time video, AI processing, and user data was an ongoing effort that continually pushed our team to innovate and adapt.

Accomplishments that we're proud of

We are incredibly proud of building a fully functional, integrated platform that delivers on our core vision. Successfully merging Tavus for video storage and behavioral analytics with our AI dialogue system was a major technical win, creating a seamless and powerful feedback loop. We built a complex data pipeline that reliably transcribes live audio, processes it for meaningful feedback on teaching skills like rapport and adaptability, and stores everything neatly in MongoDB for user review. Most importantly, we created an AI student that feels genuinely realistic and responsive, providing teachers with a valuable and safe space to practice and fail forward.

What we learned

This project was a profound learning experience. We gained deep, hands-on expertise in integrating diverse AI models and APIs, managing real-time data flows, and architecting a full-stack application with Next.js and MongoDB. We learned how to engineer effective prompts to shape consistent AI personalities and behaviors. Beyond the code, we learned about the nuances of pedagogy—what makes a teaching moment effective and how to translate that into measurable feedback. Perhaps the biggest lesson was in building a complex, multi-layered system where video, audio, AI, and data storage all had to work in perfect harmony.