EngageAI

🎬 EngageAI

Full-stack assistant for YouTube creators

💡 Inspiration

We wanted to make life easier for YouTube creators who constantly juggle analytics, comments, and editing. Our goal was to build an assistant that automates the repetitive parts of running a channel — replying to comments, spotting what content works, and generating Shorts — while keeping creators in control of every step.

⚙️ How We Built It

EngageAI combines a Node.js + Express + MongoDB backend with a React (Vite) frontend.

Backend:
- Google APIs SDK connects to the YouTube Data + Analytics APIs.
- OpenAI SDK generates comment reply drafts and script ideas.
- ffmpeg and ytdl-core power the Shorts clipping workflow.
- MongoDB stores sessions, user data, and temporary video files (GridFS).
Frontend:
- React + TailwindCSS dashboard for comments, analytics, and Shorts workflows.
- OAuth2 authentication through Google to securely link the creator’s channel.

This architecture lets creators log in, fetch channel data, generate replies or clips, and view performance analytics all in one place.

🧠 What We Learned

How to set up and use the YouTube Data and Analytics APIs effectively.
How to integrate OAuth2 authentication and handle refresh tokens safely.
How to use OpenAI for tone-aware, human-like comment replies.
How to combine multiple AI tools (OpenAI + ffmpeg) in one workflow.
The importance of building draft-first workflows so creators remain in control.

🚧 Challenges

Managing the full OAuth flow within the 36-hour timeframe.
Handling API quotas and rate limits during testing.
Debugging ffmpeg and file storage issues when trimming Shorts.
Keeping session tokens persistent across backend and frontend environments.

🚀 What’s Next

Deeper video insights using the YouTube Analytics API (retention, CTR, audience).
AI-assisted thumbnail ranking and A/B testing.
Multi-platform support (TikTok, Instagram Reels).
Deployment on a scalable cloud environment with CI/CD.