Inspiration

We wanted to create a way for users to see their “year in TikTok” in a wrapped format—analyzing their viewing patterns, behaviors, and engagement, while also generating a fun personality type based on what they watch.

What it does

Our service takes in user watch history and behavior data, sends it to a trained DigitalOcean Gradient AI agent for analysis, and returns insights, statistics, and a personality profile presented in a fun, engaging way.

How we built it

Backend: Flask app handles file uploads, data formatting, and AI agent integration.

AI Analysis: DigitalOcean Gradient agents process the metrics and generate insights, personas, and humorous commentary.

Frontend: Simple and interactive interface displaying statistics, top categories, emotional patterns, and AI-generated personality summaries.

Challenges we ran into

  • Connecting to DigitalOcean Gradient agents: Setting up private endpoints and handling access keys for secure communication was tricky.
  • Parsing and normalizing CSV data from various sources to feed into the AI agents reliably.
  • Ensuring fast response times while orchestrating multiple agents for metrics, persona, and humor generation.

Accomplishments that we're proud of

  • Successfully integrated multiple AI agents to generate detailed, personalized insights.
  • Created an engaging and visually satisfying “wrapped” experience with meaningful metrics.
  • Developed a flexible backend that can be extended to handle multi-agent orchestration and future personalization features.

What we learned

  • How to securely connect and authenticate with private AI agents on DigitalOcean.
  • Effective techniques for transforming raw user data into actionable metrics.
  • Orchestrating multiple AI agents to produce a coherent, entertaining, and informative output.

What's next for Deepfeed Wrapped

  • Add features that allow judges or users to input preferences and generate custom analyses.
  • Expand multi-platform support, especially mobile, for better accessibility.
  • Automate data generation from scraping or other sources to simulate user activity without relying solely on uploads.
Share this project:

Updates