MonBac

• 

• 

  home page

• 

  ai chat console

• 

  ai chatting with the user

Inspiration High school and college students frequently struggle when studying independently. While standard AI chatbots exist, they often lack curriculum context and are prone to hallucinations, confusing students further. I wanted to build a platform where the AI knows exactly what the student is currently reading and can provide laser-focused, accurate tutoring in real-time.

What it does MonBac is a full-stack Learning Management System tailored for students. Its core feature is the integrated AI Course Tutor. When a student is inside a specific lesson (e.g., studying a specific math or science concept), the platform invisibly packages the current chapter and lesson metadata. It sends this enriched context alongside the student's question to the AI, ensuring the response is strictly relevant to the curriculum they are looking at.

How we built it

AI Infrastructure: I utilized the Nebius AI Cloud Serverless Endpoints, specifically querying the nvidia/Nemotron-3_5-Lightning model. I chose the Lightning model because sub-second latency is critical for a conversational tutoring interface.

Backend: Built with Node.js and Express, utilizing the official OpenAI Node SDK to route prompts directly to the Nebius API.

Database: Migrated from a local instance to an Aiven Cloud MySQL database, implementing strict SSL connection rules.

Frontend: Embedded JavaScript (EJS), HTML5, and CSS3, engineered to fully support both LTR (French) and RTL (Arabic) interfaces for Moroccan students.

Deployment: The live production environment is hosted on Render.

Challenges we ran into Migrating the local database schema to a secure cloud database in the final hours of the hackathon was a major hurdle. I had to resolve legacy MySQL terminal authentication issues and rewrite the database connection pool (db.js) to strictly enforce SSL configurations required by Aiven, all while ensuring the production server on Render could safely access environment variables without hardcoded secrets.

Accomplishments that we're proud of Successfully engineering the context-injection pipeline. Getting the AI to understand exactly which lesson the student is on—without the student having to explain it themselves—makes the platform feel like magic. I am also extremely proud of deploying a highly secure, production-ready full-stack application under intense time pressure.

What's next for MonBac Expanding the platform to support voice interactions so students can talk to the Nemotron model natively, and integrating an AI-driven quiz generator that creates personalized assessments based on the student's chat history with the tutor.

Built With

• aiven
• and-css3
• ejs
• express.js
• html5
• implementing-strict-ssl-connection-rules.-frontend:-embedded-javascript-(ejs)
• javascript
• mysql
• nebius
• node.js
• nvidia
• render
• specifically-querying-the-nvidia/nemotron-3-5-lightning-model.-i-chose-the-lightning-model-because-sub-second-latency-is-critical-for-a-conversational-tutoring-interface.-backend:-built-with-node.js-and-express
• utilizing-the-official-openai-node-sdk-to-route-prompts-directly-to-the-nebius-api.-database:-migrated-from-a-local-instance-to-an-aiven-cloud-mysql-database