Research
Our published research and ongoing work across AI, domain specialization and physical intelligence.
We examine how specialized professional knowledge can be incorporated into large language models to produce more accurate, contextually appropriate responses in high-stakes domains including construction, medicine and engineering.
This paper presents our approach to interpreting architectural and engineering drawings using vision-language models, with findings on accuracy across different document types and drawing conventions.
We explore the problem of maintaining user context as AI systems transition between specialized domains, presenting a framework for coherent multi-domain interactions.
Health information represents a particularly sensitive application of AI. This paper outlines our approach to accuracy, epistemic humility and appropriate boundary-setting in the Moldulus Health intelligence.
An early-stage exploration of how AI systems can begin to perceive and respond to physical environments, with initial findings from our lab research program.
We present Moldulus Fashion as a case study in applying AI reasoning to creative and material design problems, with an analysis of how structured knowledge about material, construction and form supports the creative process.