SmartScheduler.
Complex plans. Clear decisions.
Turning a genetic algorithm’s output into a timetable people can trust.
Show the reasoning behind the timetable.
University timetables must satisfy room capacity, instructor availability and time constraints together.
My goal was to present the genetic algorithm’s schedules through an interface that people can understand and trust.
Turning complex logic into an understandable interface.
I worked on the Next.js interface, weekly grid, what-if scenarios and typed API layer.
A simple flow. Clear feedback.
The selected room is not available at this time.
Useful tools for more informed planning.
Conflict-free schedule
A weekly timetable that satisfies the constraints.
Scenario comparison
Compare alternative schedules side by side.
Export
Export the timetable as a shareable file.
“The interface must show the reasoning behind the decision.”
People need to see how the result was reached. Visible reasoning makes an algorithmically generated timetable easier to trust.
Tech stack.
Next.js 14
TypeScript
Tailwind CSS
ASP.NET Core 9
PostgreSQL 16
Docker
GitHub Actions
Technical notes and project scope
The problem and my role
A university timetable must satisfy room capacity, faculty availability, student cohorts and time constraints all at once. Done by hand, clashes are inevitable. Our team built a genetic-algorithm engine to solve it — but the raw output of an optimisation engine is useless to anyone on its own.
I am Frontend Lead in a five-person Scrum team. The genetic-algorithm engine and the .NET services are not my work — that side belongs to another member of the team. My responsibility was making the engine’s solution something a human can read and interrogate.
Implementation decisions
- I visualised the solution on a weekly grid: which class sits where, which constraint put it there, which alternatives were eliminated.
- I added what-if scenarios — when a user tries to move a class, they immediately see which constraint breaks.
- I built the interface with Next.js 14, TypeScript and Tailwind, pinning the contract with the backend through a typed API layer.
Outcome and learning
Clash-free generation, scenario comparison and export all work. The project ran on Docker with GitHub Actions continuous integration.
The hardest part of an optimisation system is not the optimisation — it is persuasion. If a user cannot see why a class was placed where it was, they will override it by hand no matter how good the algorithm is. The interface’s job is not to show the answer but to show the reasoning.