TLE Doctor

Inspiration

I solved it. The judge still said TLE. At the ICPC Dhaka Regional, my logic was correct — but the verdict was Time Limit Exceeded. The judge tells you nothing: not why it's slow, not where, not how to fix it. Forums take hours. Tutorials are generic — never about YOUR code. So beginners keep guessing, and keep getting TLE. Speaker note: "This happened to me. That frustration is why this project exists."

What it does

TLE Doctor — Diagnose. Treat. Cure. Paste your code → get a full medical-style diagnosis in seconds. 8-section report: verdict PASS / RISKY / TLE, complexity, bottleneck, prescription with fixed code. Plus: progressive hints (3 levels), a TLE Patterns Library, a Complexity Calculator, and an Example Gallery.

How it works

Paste code → Doctor examines it → You get the cure [Screenshot of the live app + a sample report here]

1. Paste code, pick the language. 2. Press "Diagnose my code". 3. Read the verdict, the slow loops, and the fixed code. Built with HTML, CSS, JavaScript + Google Gemini API. One file, hosted on Netlify — no server, no account. Your free Gemini key stays in your own browser only. No key? The sample diagnosis, gallery, patterns and calculator still work. Speaker note: "One AI call, one clear report. A judge can test the whole product without signing up for anything."

Challenges we ran into

1. Getting the AI to answer in the same clear format every time — solved with structured prompt engineering: a fixed 8-section medical report the AI must follow.
2. My first live test failed with a "model not found" error — the AI models I had listed had been retired by Google. I fixed it by making the app automatically discover which models are actually available for the user's key, so it can never break that way again.
3. Making it beginner-friendly — simple English, real numbers (10¹⁰ operations vs the ~10⁸ per second a judge allows), and a doctor metaphor instead of scary jargon.
4. Making the demo bulletproof — a built-in sample diagnosis that works even without an API key, so anyone can test the full product in 10 seconds.

Accomplishments that we're proud of

1. Built and shipped a working AI product as a first-time hackathon participant, completely solo
2. Turned my own ICPC TLE failure into a tool that helps other students
3. A memorable "medical diagnosis" theme that makes complexity analysis feel friendly, not scary 4.A full product, not just one feature: diagnosis, progressive hints, a patterns library, a complexity calculator and an example gallery — completely free for any student to use

What we learned

1. Prompt engineering decides the quality and consistency of an AI product
2. User experience matters as much as the technology — a judge understands your product in 2 minutes, or never
3. Personal pain points make the best projects
4. In a hackathon, a small finished product beats a big unfinished one

What's next for TLE Doctor

What's next — and never TLE again Shareable report links + saved diagnosis history. Real code execution and benchmarking against the clock. Browser extension for Codeforces / AtCoder. Closing line: "Diagnose. Treat. Cure. Never TLE again." — tle-doctor.netlify.app Speaker note: "The product works today. These are the next steps after the hackathon."

Built With

• css
• google-gemini-api
• html
• javascript