Mileo

Inspiration

I spend about 4 hours a day commuting to my internship and back. That's time I can't use for deep work, but I kept thinking it could be enough for something small: a few flashcards, a quick review. I built Mileo to turn dead commute time into short, focused study sessions.

What I learned

• Integrating the Gemini API to generate flashcard decks from any topic or uploaded notes (PDF/DOCX/TXT)
• Persisting sessions with SQLite so progress survives refreshes
• Handling weekly stats correctly across timezones (Mon-Sun, Asia/Kolkata)
• Writing server-side validation and a real test suite with pytest

How I built it

The backend is Python + Flask, the frontend is server-rendered HTML with plain JavaScript and CSS, and sessions are stored in SQLite. Decks are generated live with the Gemini API (gemini-2.5-flash) from a topic and difficulty, or from uploaded study material. The dashboard tracks weekly minutes, hours reclaimed, and recent sessions.

Challenges

• Preventing duplicate session saves from double taps
• Getting week boundaries right across timezones
• Making AI-generated decks reliable enough to study from
• Building this alongside a full-time internship and a 4-hour commute - which is exactly the problem Mileo tries to solve.

Built With

• api
• css
• edtech
• education
• flashcards
• flask
• generative-ai
• google-gemini-api
• html
• javascript
• microlearning
• mobile-first-design
• pypdf
• pytest
• python
• python-docx
• python-dotenv
• rest
• sqlite
• web-app