EOD Personalized

We've all been there: It's 11 PM, you have 50 tabs open, and you've "learned" everything from sourdough baking to quantum mechanics... or have you? Streamline your internet activity towards creativity

Comment

EOD Personalized Preview

EOD Personalized: Turning Digital Exhaust into Digital Growth 🚀

Try it out Live!

Inspiration 💡

We've all been there: It's 11 PM, you have 50 tabs open, and you've "learned" everything from sourdough baking to quantum mechanics... or have you? We realized that our browsing history isn't just a list of sites—it's a goldmine of latent intent. We wanted to build something that stops the doom-scrolling and starts the actual learning. EOD Personalized was born from the desire to take that messy "digital exhaust" and refine it into a crystal-clear roadmap for personal growth.

What it does 🤖

EOD Personalized is your AI-powered growth companion. It takes your raw Google Takeout data (History, Bookmarks, Reading List) and transforms it into actionable insights.

  • 🔍 X-Ray Vision for Your History: Uses Gemini 3 Flash Preview to analyze thousands of URLs, cutting through the noise to find your true interests.
  • 🗺️ Interactive Learning Journey: Visualizes your evolving interests over time with a custom D3.js timeline—watch your "React" phase turn into a "Next.js" phase.
  • 📅 One-Click Focus: Automatically generates a 7-Day Study Plan based on what you should learn next and syncs it directly to your Google Calendar.
  • 📚 Deep Dive Reports: Compiles a curated "textbook" of your best found resources and saves it to Google Drive, ready for your next deep work session.

How we built it ⚙️

We went full stack and bleeding edge:

  • The Brains: Google Gemini 3 Flash via the GenAI SDK. We pushed the context window to the limit, feeding it structured history data to get precise JSON outputs.
  • The Body: Next.js 15 (Pages Router) + HeroUI (v2). We didn't just want it to work; we wanted it to feel premium. Glassmorphism, framer-motion animations, and responsive layouts.
  • The Engine: Python Flask. Handling massive ZIP file uploads, parsing HTML/JSON in memory, and acting as the orchestration layer between the frontend and the LLM.
  • The Visuals: Custom D3.js implementation for the timeline component—because standard charting libraries couldn't capture the "journey" vibe we wanted.
  • Integrations: LangChain for YouTube search tools, plus deeper Google Workspace API integration (Drive & Calendar).

Challenges we ran into 🏔️

  • Taming the LLM: Getting consistent, deeply nested JSON for our complex data models (Timelines, Categories, Daily Plans) from a creative model was a battle. We used rigorous Pydantic definitions and iterative prompt engineering to lock it down.
  • Data Overload: Parsing 100MB+ user history files in-memory without crashing the server required some serious optimization and smart filtering heuristics.
  • The "Perfect" Timeline: visualising abstract "topics" over time is hard. We spent hours tweaking D3 forces and scales to make the timeline feel organic yet informative.

Accomplishments that we're proud of 🏆

  • The "Magic Button": Hitting "Sync to Calendar" and watching a messy week of browsing turn into a structured learning schedule on Google Calendar.
  • Privacy First: We built it so you can analyze your local ZIP file without it ever leaving your machine (mostly) or requiring complex auth if you don't want strict integrations.
  • Aesthetic Quality: We tried to keep the UI as simple as possible so that any user can use it without any confusion. The HeroUI implementation is crisp.

What we learned 🧠

  • Gemini is a Data Scientist: We treated Gemini less like a chatbot and more like a fuzzy logic processor. It's incredibly good at clustering vague scraping data into coherent topics.
  • Schema is King: Defining strict interfaces between our Python backend and TypeScript frontend saved us from chaos during the hackathon crunch.

Business approach/revenue model 💸

To transition EOD Personalized from a hackathon project to a sustainable product, we envision a multi-stream revenue model:

  • 🎯 Hyper-Targeted Advertising: Since the platform understands a user's specific learning journey (e.g., "Transitioning from Java to Python"), we can serve highly relevant ads for specialized dev tools, IDEs, and hosting services that convert at a much higher rate.
  • 🔗 Affiliate Learning Partnerships: When the AI suggests a "Next Topic," we can provide affiliate links to premium courses on platforms like Coursera, Udemy, or Pluralsight, earning a commission on every signup.
  • 💎 Freemium "EOD Pro": The core analysis remains free. A premium subscription would offer longitudinal tracking (multi-year history), deeper cross-platform insights (YouTube + GitHub integration), and a "Personal AI Coach" that checks in on your study plan progress.
  • 📊 B2B Insights: Selling anonymized, aggregated data on emerging developer technology trends to HR tech companies and EdTech providers to help them understand what skills the talent market is currently focusing on.

What's next for EOD Personalized 🔮

  • NotebookLM Integration: Auto-generating an "Audio Overview" podcast of your week's reading list.
  • Gamification: "Focus Streaks" to keep you accountable.
  • Multi-Platform: Pulling in GitHub commits and YouTube watch history for a 360-degree view of your developer journey.

Check out the code and try it with your own Google Takeout!

Built With

Share this project:

Updates