MentEd

Inspiration

High quality mentorship is incredibly valuable but often inaccessible due to cost and scheduling constraints. Platforms like Topmate offer great one on one sessions, but the knowledge shared in those meetings is lost after the call ends. We wanted to build a platform that compensates experts fairly while making their insights freely available to the public.

What it does

MentEd is a dual sided mentorship marketplace integrated with an AI knowledge engine. The premium layer allows users to book paid video sessions with industry mentors to solve specific problems. These sessions are automatically recorded and transcribed. For the free layer, these transcripts are vectorized and fed into our AI chatbot. Anyone can search this database for free. If the AI can answer a question, it cites the specific mentor. If it cannot, it prompts the user to book a paid session with the relevant expert. Additional features include a public Ask Me Anything board, mentor portfolios, and a trust and safety refund guarantee.

How we built it

We built the frontend and backend using Next.js and Tailwind CSS. User authentication and our PostgreSQL database are managed through Supabase. For the core video functionality, we integrated Jitsi as a Service with secure JWT backend authentication. After a session concludes, we use AssemblyAI to generate transcripts. Finally, we power the conversational chatbot using Google Gemini via the Google AI SDK to parse and retrieve answers from the transcript data.

Challenges we ran into

Securing the video infrastructure using Jitsi JWTs required careful handling of private keys and environment variables. We also faced challenges managing asynchronous webhooks for the video transcription process and ensuring our Git repository remained secure during rapid prototyping.

Accomplishments that we're proud of

We are incredibly proud of successfully connecting the entire core loop. A user can book a session, hold the video call, have the transcript generated, and immediately see that new knowledge reflected in the AI chatbot's responses.

What we learned

We learned a significant amount about integrating third party video APIs, handling secure authentication flows, and utilizing large language models for highly contextual data retrieval based on live user interactions.

What's next for MentEd

Moving forward, we plan to implement a live payment gateway like Stripe to handle real transactions and refunds. We also want to refine the AI vector search to provide even more accurate citations and improve the mentor matching algorithm.

Built With

• assemblyai
• claude
• claudecode
• gemini
• javascript
• jitsi
• next.js
• node.js
• postgresql
• react
• supabase
• tailwing.css