Eigenform AI

Eigenform is a Singapore-based AI research and engineering company developing local-first systems that learn from experience and improve over time. Its approach connects AI models to real environments, where agents can inspect information, form hypotheses, use tools, test their ideas, and learn from measurable outcomes. Successful interactions can then inform future model training, with reasoning traces designed to be verifiable.

The company builds a full-stack platform spanning software, adaptable agent harnesses, models, and ruggedised local computing hardware. Its prototype compute system combines GPU processing, storage, and a complete computer for running AI and processing data locally, including where connectivity is unreliable or information must remain under an organisation’s control.

Eigenform’s open-source Geocluster workspace supports geological exploration and earth sciences. It helps users clean and structure reports, maps, and exploration datasets; analyse geological evidence with AI agents; and visualise spatial relationships and model outputs. The company also publishes research on continual learning and recursive self-improvement, alongside case studies and experiments that examine how systems can learn from practical results.