DeadlinePilot AI

• 

  Sample data showing the schedule and at-risk logic. Tasks are generated by Gemini at runtime.

• 

  Risk detection: Physics is at risk (20h left, 1 day), Maths is safe. Sample data.

• 

  Type a deadline in plain language, and DeadlinePilot AI builds the plan.

Inspiration

Students and freelancers rarely miss deadlines because they forget them. They miss them because they don't know how big the work is or where to start. Normal to-do apps only store a due date. They don't build a plan. We wanted an assistant that turns "Maths project by 30 October, 6 hours of work" into an actual schedule and warns you before you fall behind.

What it does

DeadlinePilot AI is a smart deadline assistant:

• Understands plain language: type a deadline in everyday words (even Hinglish) and the AI extracts the title, due date, and estimated effort.
• Breaks work into subtasks: the AI splits the work into 3-6 manageable tasks with hour estimates.
• Schedules automatically: tasks are spread across the days between today and the due date.
• Tracks progress: tick off tasks and watch the progress bar fill.
• Warns about risk: a 🔴/🟢 indicator tells you if your remaining work exceeds the time you realistically have.

How we built it

• Frontend + backend: Python with Streamlit
• Database: SQLite (deadlines and tasks tables)
• AI: Google Gemini API for parsing and task breakdown
• Risk logic: plain Python, not AI. Risk is calculated as:

$$\text{at risk} \iff \text{remaining hours} > \text{days left} \times \text{hours per day}$$

We deliberately kept the math in code and used the AI only where language understanding is needed. That makes the warnings reliable.

Challenges we ran into

• Getting API access working on a restricted laptop (blocked pip.exe, so we used python -m pip).
• Handling API rate limits and temporary 503 errors on the free tier, which we solved with retry logic and by reducing the number of AI calls per request.
• Making the AI return clean JSON every time, with parsing and error handling.

What we learned

• Use AI for language, and normal code for calculations.
• Always handle API failures gracefully.
• Start with a small, working MVP and ship it.

What's next

• Telegram / WhatsApp / email reminders
• A daily AI briefing: "here is what matters today"
• Google Calendar sync
• Team deadlines and a mobile app

Built With

• google-gemini-api
• python
• sqlite
• streamlit