HiReady

💡 Inspiration

We were inspired by the "black box" of remote learning and corporate training. While instructors can see that a video was watched, they often have no idea where employees got confused or why they dropped off. We wanted to build a bridge between video content and student comprehension, using AI to give educators the same level of granular analytics that web developers have for user interfaces.

🚀 What it does

HiReady.tech is an identity-aware learning analytics platform. It uses AI video intelligence to index training materials, allowing employees to search for specific concepts and access an AI-assisted chat assistant. For trainers and HR, it provides a powerful dashboard that identifies "learning friction"—pinpointing specific video segments that cause confusion based on behavior like rewinds and pauses—while protecting individual privacy through pseudonymization.

🛠️ How we built it

We built HiReady using a robust multi-service architecture:

• Frontend: Built with React, TypeScript, and Tailwind CSS. We used Framer Motion for a fluid UI and Recharts for complex data visualization.
• Backend: A dual-server setup using Node.js/Express for user management and a Flask (Python) server for AI-heavy processing.
• AI & Storage: We integrated the Twelve Labs API for video indexing/summarization and Backboard for the AI chat assistant. Videos are securely stored and streamed via Cloudflare R2.
• Database: MongoDB Atlas handles our data storage and user pseudonymization mapping.

🚧 Challenges we ran into

The biggest hurdle was mapping asynchronous user behavior (like a "rewind" event) to specific AI-generated video concepts in real-time. We had to synchronize the frontend video state with the Twelve Labs indexing data to ensure that when a student struggled with a specific second of video, the system correctly identified the underlying topic. Configuring CORS and streaming headers for Cloudflare R2 also required significant troubleshooting to ensure seamless video seeking.

🏆 Accomplishments that we're proud of

We are incredibly proud of our Friction Detection engine. Successfully clustering users into "behavioral groups" (like "High-Replay Learners") and visualizing exactly which concepts caused those groups to struggle felt like a major win. We also managed to maintain a high level of data privacy by ensuring trainers only see aggregated, pseudonymized insights rather than individual student tracking.

📖 What we learned

We learned how to leverage multimodal AI APIs to extract meaning from unstructured video data. This hackathon pushed us to think beyond simple LLMs and explore how video embeddings and temporal indexing can be used to solve real-world educational problems. We also gained deep experience in managing complex full-stack environments involving both Node.js and Python microservices.

🔮 What's next for HiReady.tech

We plan to implement Predictive Intervention, where the AI proactively flags students at risk of failing a certification before they finish the course. We also want to expand our behavioral clustering with more advanced machine learning models to help HR teams optimize their training content based on how different departments (e.g., Engineering vs. Sales) consume information differently.

Built With

• backboard
• cloudfare
• express.js
• flask
• mongodb
• python
• railway
• react
• tailwind
• twelvelabs
• typescript
• vercel
• vite