The task

Probability theory and statistics are hard to learn from dry lecture notes: a student needs to see how a problem is solved and try it right away.

We wanted an open service that does not just give the answer, but explains the solution method.

Explaining the method, not giving away the answer

Step-by-step analysis, similar tasks for practice and solution checking.

The solution

We designed the service around three scenarios: analyzing a problem from the study guide with examples, generating a similar problem with different numbers or context, and checking the student's solution step by step. The neural network is connected as a user-parametrizable agent through an OpenAPI-compatible interface.

Task analysis using the study guide with examples
Task generation by types, topics, and kinds
Answer checking and step-by-step solutions
User-parametrizable neural network (OpenAPI-compatible agent)

Our contribution

Designed the service on a modern architecture and implemented the MVP: frontend, backend, integration of a user-parametrizable neural network via an OpenAPI-compatible agent, guest mode, and secure storage of user data. The project is free and open, created on the initiative of Script Agency for Perm State University.

The result

The MVP is ready and undergoing beta testing: the service is available to students and faculty of Perm State University, and the architecture allows the course to grow without reworking the interface.