StudyFlow AI

Inspiration

Every semester I watched classmates — and myself — drown in scattered schoolwork: assignments in one app, exam dates on a whiteboard, project deadlines in chat messages. We always knew what needed doing, but never what to do first, how long to spend, or whether we were falling behind. Existing to-do apps just stare back at you. I wanted to build the thing I actually needed: something that looks at the chaos and tells me exactly what to do next.

What it does

StudyFlow AI turns a student's workload into an Adaptive Academic Planner:

• Focus Now card — the single most important task right now, with plain-English reasoning ("High priority + deadline approaching + effort is high")
• "What should I do now?" — one click, one concrete next action with WHY and TIME. The signature feature.
• Deadline Radar — scans every task and flags Safe / Watch / At Risk before panic sets in
• Rescue My Week — one click rebalances an overloaded week, splits big projects into sessions, and protects Friday evening, with a before/after view
• "I couldn't finish this" — the schedule automatically adapts instead of guilt-tripping you
• Focus mode — distraction-free timer where Easy/Normal/Hard feedback improves future time estimates
• Analytics + AI Weekly Review — supportive insights, never judgmental
• StudyFlow Assistant — answers questions about your actual schedule ("Why is Biology my highest priority?")

The priority engine fuses deadline urgency, remaining effort, difficulty, importance, and workload pressure:

$$score = urgency(deadline) + effort + difficulty + importance + pressure$$

so a 3-hour hard project due in 3 days correctly outranks a 20-minute easy worksheet due tomorrow — it knows the big one needs to start now.

How I built it

Solo build. Frontend: React + TypeScript + Vite + Tailwind CSS, Framer Motion animations, Lucide icons, React Router (9 pages: landing, login, onboarding, dashboard, tasks, calendar, focus, analytics, settings). Backend: Python FastAPI with Pydantic validation and a clean REST API. Database: PostgreSQL schema (Supabase-ready), seeded with realistic demo data. AI: an AIService abstraction over any OpenAI-compatible API with the key kept server-side — plus a deterministic planning engine fallback so the demo never breaks and never shows an error to judges. Deployed frontend on Vercel, source on GitHub.

Challenges

The hardest problem was the priority math: deadline-only sorting kept recommending quick wins while big projects quietly became emergencies. I iterated the scoring weights until a hard 3-hour project due in 3 days correctly outranked an easy task due tomorrow. Second challenge: making every AI feature work with zero API key, so I built a deterministic engine that mirrors the AI outputs and labeled it "demo intelligence mode." Finally, the judge-first 60 seconds — the dashboard had to tell the whole product story instantly, which took three redesign passes toward a calm Linear/Notion-style aesthetic with light + dark mode.

Accomplishments

A complete, deployed, working product — not a mockup. Every button works: 27/27 backend API checks passing, all pages rendering with zero console errors, verified with headless-Chrome click-through tests. The "Rescue My Week" before/after moment genuinely feels like magic.

What I learned

I learned that good student productivity isn't about more lists — it's about reducing decisions. Technically, I leveled up in FastAPI/Pydantic API design,-score based ranking systems, SPA deployment with rewrite rules, and testing real user flows headlessly. strategically, I learned to design for the first 60 seconds.

What's next

Render + Supabase deployment for AI-powered mode, real Supabase Auth, drag-and-drop prioritization, class-schedule import, spaced-repetition exam prep, and push reminders before Radar risks turn red.

Built With

• fastapi
• framer-motion
• lucide
• openai
• postgresql
• pydantic
• python
• react
• react-router
• supabase
• tailwind-css
• typescript
• uvicorn
• vercel
• vite