This research introduces a framework for creating interactive AI systems that assist scientists in solving complex problems through structured program induction.

• Proposes a software engineering approach for building scientific assistants
• Leverages large language models with interactive human guidance
• Focuses on practical implementation rather than specific scientific problems
• Aims to accelerate solutions for urgent scientific challenges

This work matters for engineering because it bridges the gap between AI capabilities and scientific applications, providing a systematic methodology for developing tools that augment human scientific expertise rather than replacing it.

Engineering Scientific Assistants using Interactive Structured Induction of Programs