CMU Bites: When My Deadlines Started Influencing My Dinner
- Indu Arimilli
- Aug 15
- 2 min read
Updated: Oct 8
By Indu Arimilli | August 2025
It’s 11:47 PM. I have a project due in 13 minutes. My choices: instant ramen, panic, or both.
That’s when it hit me — my food habits were basically a reflection of my stress cycle.So, I built CMU Bites, an AI that reads your mood, deadlines, and vibe… and decides what you should eat.
Research & Discovery
This idea started as a casual survey I ran across CMU group chats:
“What do you eat when you’re stressed?”
30 students responded.68% said their food decisions depend on “energy level,” not nutrition. Several mentioned how dining apps “never get the context” — recommending sushi when you’re clearly in existential crisis mode.
Problem Statement:
Students make reactive, stress-based food choices because current recommendation systems ignore mood and workload context.
Goal: Design a recommender that personalizes meals not just by taste, but by stress.
MVP: CMU Bites
An AI-driven mood-to-meal recommender that syncs your calendar, sentiment, and CMU dining data.
Core Features:
Mood Detection: Infers stress level from messages, deadlines, or calendar load.
Recommendation Engine: Cross-analyzes mock dining reviews, sentiment tone, and nutritional data.
Chat UI: Speaks like your overworked friend. Example:
“You sound sleep-deprived. May I suggest a bagel and an iced coffee?”
User Story:
“As a student juggling deadlines, I want food suggestions that fit my mood and schedule so I can eat better (and maybe survive finals).”
Success Metric:
Reduce “indecision time” before ordering by 30% (measured through simulated test flows).
The Prompt
“You are an AI campus dining assistant named CMU Bites.Using mock data of dining options (name, cuisine, calories, and student sentiment), build a recommender that:
Analyzes text inputs like ‘I’m exhausted but starving’ or ‘I just finished an exam.’
Classifies the emotional state (e.g., stressed, sleep-deprived, celebratory).
Maps that emotion to appropriate food categories (e.g., comfort, energy-boosting, indulgent).
Returns three meal suggestions with witty, human-like messages and emojis.
Format output as a chat message for a mobile interface.”
The Visual
My Takeaway
Personalization isn’t just about preferences. It’s about context — because no one craves kale when their group project just crashed five minutes before the deadline.