SortingHat AI

Smarter room allocations for RCs

Comment

WORKFLOW

  1. Admin sends a unique code to the student
  2. Student enters the code into the quiz website and does the quiz
  3. Data is automatically collected and stored in HR DB
  4. RC HR generates optimal room allocation
  5. Allocation finalised
  6. Student gets an alert to check the website, get their house and room allocation.
  7. After the semester ends, obtain feedback on student thoughts about their level allocation and train model based on feedback loop

Inspiration

  • Inspired by the Harry Potter movie series, concept behind having the sorting hat to assign students to their respective houses

What it does

SortingHat AI replaces the manual spreadsheet-based room allocation process used by residential colleges with an AI-powered system. Incoming students complete a short personality quiz based on the Big Five framework, and our algorithm optimises floor assignments to maximise diversity across faculty, nationality, personality type, and interests while respecting practical constraints like accessibility needs, gender preferences, and sleep schedules. An HR dashboard gives housing staff full visibility and override control, turning weeks of manual sorting into hours of reviewing smart recommendations.

How we built it

Nextjs for the full-stack for student quiz + HR dashboard, with Supabase handling auth and database. GPT-4o for measuring the open-ended quiz questions to the correct OCEAN archetype we also used MILP (Mixed Integer Linear Programming) for the optimisation room allocation

Challenges we ran into

Deciding on the optimal room allocation algorithm

Accomplishments that we're proud of

We're proud that the system genuinely works end-to-end, from a student completing the quiz to the algorithm producing a diversity-optimised allocation that HR can review and override in real time.

What we learned

  • It was an interesting experience for us to work on this initiative, having conducted interviews with RC administrative managers and students to understand their painpoints, as we worked together as a team to build this.

What's next for SortingHat AI

  • By Round 2, we will have a fully working MVP and every stage of the workflow will be shown.
  • With approval from residential services team, we plan to trial it in one residential college during the upcoming room allocation exercise for AY2026/27 in July 2026. Following reviews on efficiency, we are looking to roll out this initiative across more residential colleges in the following cycles after fine-tuning of the software.

Built With

  • milp
  • nextjs
  • supabase
Share this project:

Updates