Research
Research that becomes responsible products.
We study how models, people, interfaces and enterprise systems work together under real constraints.
Agents
Enterprise agents with bounded authority
Tool use, permissions, escalation and human review.
Knowledge
Retrieval quality and evidence visibility
Permission-aware search, citations and evaluation.
Evaluation
Testing probabilistic systems
Realistic task evaluation before and after deployment.
Multimodal
Language, vision and speech
Complete workflows that combine multiple forms of information.
Human–AI
Interfaces that support judgment
Controls, sources and escalation designed around people.
Security
Secure model and tool orchestration
Least privilege, auditability and defense against misuse.