Tech Stack
Description
This project was built around a practical problem in Alloy: once the solver starts generating many valid instances, it becomes difficult to find the ones that are actually useful, diverse, or worth inspecting. The project explores the idea of 'interestingness' as a ranking and selection problem over valid generated instances.
I worked on turning that problem into something interactive and explainable by combining scoring logic, diversity-based picking strategies, and a UI layer for exploring generated solutions. The project sits more in formal methods, algorithms, and system tooling than in classical AI, but it still shows strong reasoning and ranking-system design.
- Built an interestingness-ranking workflow on top of Alloy-generated instances
- Worked on tuple-score and diversity-based selection strategies to reduce 'too many instances' overload
- Implemented Java-based UI and evaluation flow for exploring and comparing generated solutions
- Connected Alloy, Kodkod, and solver-generated outputs into a more usable inspection workflow
- Turned a formal methods problem into a practical interactive system for instance analysis
Page Info
Interestingness Explorer UI
UI for viewing generated instances, scores, diversity signals, and ranked selections.
Selection & Analysis Results
Comparison of ranked and diversity-based instance selection strategies over Alloy solutions.
