From Natural Language to Interactive STEM Simulations

April 15, 2026

Virtual laboratories are useful only when students can move from an idea to an experiment without fighting the interface. Ingenium's text-to-simulation work explored how natural-language and voice instructions could become interactive STEM experiences.

A student might describe a physics setup in ordinary language. The system must identify the relevant objects, parameters, constraints, and expected interactions before generating a usable simulation state. That requires more than a single model response: it needs structured interpretation, validation, rendering logic, and clear feedback when a request is incomplete.

The product loop

  1. Interpret the learner's intent.
  2. Convert it into a structured experiment specification.
  3. Validate values and physical constraints.
  4. Render the experiment through a WebGL-based interface.
  5. Use an adaptive tutor and learning telemetry to guide the next step.

The project brought together NLP, full-stack engineering, interactive graphics, and education-product design. It was developed through the University of Sydney INCUBATE ecosystem and became a foundation for both startup experimentation and research into natural-language interfaces for intelligent STEM simulation.

GitHub
LinkedIn