TartanTempo

Inspiration

TartanTempo was inspired by the daily chaos most CMU students face. With dense class schedules, irregular dining hall hours, and limited time for fitness, it can feel impossible to plan meals or workouts consistently. We wanted to create something that quietly handled all of that planning in the background and gave students a healthier, more efficient routine.

What it does

TartanTempo lets students upload their .ics class schedule, enter basic nutrition and fitness preferences, and automatically generates an optimized weekly plan for meals and workouts. Using Claude AI, it aligns free time, dining hall availability, and gym hours to create a realistic schedule that students can preview and export directly back into their calendar.

How we built it

We built a lightweight React web app that supports ICS uploading, a minimal preference survey, and a dynamic schedule preview. The backend parses the student’s calendar, gathers dining and recreation hours, and identifies free blocks of time. These structured inputs are sent to Claude with a strict JSON schema, and its response is rendered in the UI. We then convert the final plan into an .ics file that students can download and use immediately.

Challenges we ran into

Throughout development we struggled with parsing inconsistent ICS formats, pulling reliable hours from CMU dining and recreation sources, and ensuring that Claude consistently returned valid JSON under time pressure. Designing a schedule that felt human and not robotic was also difficult, especially given the short hackathon timeline. Getting all parts of the pipeline to work smoothly in under an hour was one of the biggest hurdles.

Accomplishments that we're proud of

We are proud that we built a complete end-to-end workflow in such a short window: upload, preferences, AI generation, preview, and export all function seamlessly. The schedules Claude generated were surprisingly usable and human-like, and we were able to design a clean, simple UI despite the time constraints. Transforming loose preferences into structured, machine-readable constraints was also a major win.

What we learned

We learned how important tightly controlled prompting is for consistent AI reasoning. Using detailed schemas dramatically improved Claude’s reliability. We also saw how a few simple user preferences meaningfully change scheduling outcomes, and how much can be done with ICS data once it is properly parsed. The project reinforced the value of constraint-based design and thoughtful data modeling when working with AI agents.

What's next for TartanTempo

Next, we want to support meal suggestions based on dining hall menus and nutrition goals, integrate wearable data for personalized energy profiles, and add features for coordinating schedules among groups of friends. We also plan to build a mobile version with real-time adjustments and smarter optimization that accounts for crowd levels, travel time, and routine changes.

Built With

• claude
• css
• ical.js
• next.js
• react18
• tailwind