MentorVerse

MentorVerse

Inspiration

Last month, we both got rejected from three online assessments in a row. Each time, we felt more defeated and unsure of what went wrong. On a long flight afterward, my teammate texted me: 'Let's build something that fixes this.'

By the time we met up, we spent the entire drive to Penn State brainstorming what became MentorVerse—an AI-powered interview platform that helps people practice smarter, not harder.

We also realized how few resources exist for neurodivergent individuals preparing for interviews, so we decided to build a space where everyone, regardless of their learning style or communication challenges, can practice in a supportive, adaptive environment.

What it does

Email Parsing, allowing users to upload or paste interview invitation emails, automatically extracting the company, position, and specific requirements. Personalized Practice, where AI generates mock questions based on the user’s role, experience level, and interview type. Behavioral Analysis, providing real-time feedback on posture, eye contact, and tone using OpenCV-based camera analysis. AI Coach, a voice-enabled mentor powered by OpenAI that delivers live feedback, context-aware guidance, and adaptive suggestions during practice. Progress Dashboard, which tracks improvements across confidence, clarity, and consistency over time, giving users actionable insights to refine their performance.

How we built it

We designed MentorVerse with a dual focus — realism and accessibility — ensuring that every feature mimics real interviews while remaining intuitive for all users.

Frontend: Built with Next.js 16, TypeScript, and Tailwind CSS, enhanced with Radix UI components and Framer Motion for smooth, responsive animations. Backend: Developed using FastAPI (Python), with Google API powering question generation and conversational intelligence. ElevenLabs – provided natural, mentor-like voice interactions, integrated via OpenRouter for real-time AI coaching and feedback. LLM (HuggingFace, fine-tuned) – analyzed user movements and behavior to provide personalized guidance and improve performance in behavioral interview practice. Gemini API – handled parsing of interview invitation emails and sending structured data to the backend for personalized practice session generation. Gemini 2.5 Computer **– enabled personalized, automated guidance for neurodivergent users, bringing **MentorVerse closer to fully adaptive AI-powered interview coaching. Data Validation: Used Pydantic to manage structured data and ensure secure communication between services.

Challenges we ran into

We faced challenges in multi-system integration, connecting the FastAPI backend with the Next.js frontend while keeping responses consistently low-latency. Ensuring real-time feedback was another major challenge, as processing live camera input for the behavioral module without lag required intensive optimization and tuning. Managing email parsing variability proved difficult due to inconsistent formats in interview invitation emails, demanding adaptive parsing and validation logic. Lastly, designing an inclusive UX for neurodivergent accessibility, focusing on users with ADHD, autism, and anxiety, pushed us to rethink conventional UI patterns and emphasize empathy-first design.

Accomplishments that we're proud of

We built an end-to-end AI system that feels like a real mentor, not just another chatbot. We successfully combined behavioral analysis, email intelligence, and AI coaching into one seamless and intuitive experience. We designed accessibility-first features tailored for neurodivergent learners, a space that is rarely addressed in traditional interview preparation tools. We also delivered a privacy-conscious setup where user sessions remain local, keeping all personal data secure and confidential.

What we learned

We realized that the difference between an AI assistant and an AI mentor lies in empathy. Building MentorVerse taught us that truly effective feedback should encourage and motivate users, not intimidate them. To make the platform inclusive, we redesigned our entire UI and UX with accessibility in mind. We wanted neurodivergent users to feel supported, confident, and understood throughout their practice sessions. We also discovered that running real-time AI systems requires optimization at every level, from data validation and API coordination to voice synthesis and video analysis. After many iterations, tests, and late-night debugging sessions, we built a workflow that feels human, supportive, and adaptive, exactly what we wished we had during our own interview preparation.

What's next for MentorVerse

Resume & Cover Letter Review: Get instant, AI-powered feedback to optimize your application documents and pass the initial screening. Multi-language Support: Practice interviews and receive coaching in your native language for true global accessibility. Neurodivergent Practice Modes: Use specialized settings to comfortably prepare for interviews by accounting for challenges like ADHD and social anxiety. Mentor Matching: Connect with real, experienced human mentors when you need personalized guidance beyond the AI. Mobile App: Practice your interview skills anytime, anywhere with a dedicated app for your phone.where.