These flagship platforms bring together mission, capability, and translational pathways across intelligence, science, and health.
Each platform combines live systems work, open problems, translational pathways, and opportunities for students and collaborators.
This platform asks what it would take to build reliable AI systems when the missing layer is not basic capability, but agency, critical thinking, judgement, and taste, and when deeper questions about intelligence itself are back on the table.
Develop AI systems that can learn, reason, code, use tools, maintain context, and still remain directed by human agency, critical thinking, and judgement across practical scenarios.
FocusMany baseline capabilities are now widely available. The opportunity is to move from impressive demos to dependable systems that make room for agency, critical thinking, judgement, trustworthy action, and new foundations for intelligence beyond scale alone.
Agent architectures, decision quality, evaluation, coordination, trustworthy behaviour, the physics of intelligence, low-energy AI, embodied cognition, and alternatives beyond standard Transformer-era assumptions.
Live work now includes the Ark expert-agent ecosystem, Eight-Sense embodiment, operator-first architecture research, energy-aware AI through 20W AI, theory-building around the physics of intelligence, and Feynman as a hybrid quantum-classical systems architect for circuit design, verification, and advantage auditing.
Legacy platform archiveIdeal for students and collaborators interested in agency, judgement, alignment, efficient intelligence, embodiment, and the design of systems that complement human intelligence rather than merely imitate it.
This platform focuses on agents and systems that can think scientifically, navigate large discovery spaces, and partner with human experts.
Automate expensive discovery loops and bring scientific structure into AI systems, from literature grounding and research planning to experiment selection, analysis, and explanation.
FocusGenerative models, tool use, scientific datasets, and coding agents create a moment to build systems that accelerate discovery rather than merely summarise existing knowledge.
Talks on the fifth paradigm of discovery, AI for materials, and scientific discovery as a frontier capability program.
Legacy platform archiveCurrent systems include Einstein for scientific discovery, Gauss for AI-assisted mathematics, and workflow patterns that connect ideation, experiment design, theorem handoff, and verification.
Best suited to researchers in materials, chemistry, computational biology, mathematics, and scientific infrastructure who want to build new systems together.
The emphasis is on systems that translate into better care, stronger monitoring, and better health understanding while fitting actual clinical practice.
Develop scalable and aligned AI systems that translate into clinically useful tools and inform health policy, especially where complexity, workflow friction, and resource constraints matter.
FocusHealthcare now needs systems that can handle multimodal evidence, longitudinal context, regulation, workflow friction, and deployment reality rather than isolated benchmark tasks.
Projects spanning cerebral palsy assessment, mental health, Covid-19 response, imaging, longitudinal modelling, and new evidence-centric directions such as MRI report generation beyond plain autoregressive decoding.
Legacy platform archiveCollaborations with clinicians, hospitals, health researchers, and government teams working on high-impact problems with deployment potential.
See the project archive and health-specific materials for deeper history and examples.
Health projectsResearch here is increasingly supported by internal systems for discovery, invention, theorem proving, supervision, and funding strategy.
Einstein is being developed as a scientific discovery engine with rediscovery benchmarks, operator libraries, contradiction graphs, and constructive-science workflows.
Virtual PhD and professor tools are being used to rethink supervision, literature analysis, idea generation, experiment planning, and project judgement in the AI era.
GrantOS and related research strategy tooling support proposal design, reviewer simulation, opportunity diagnosis, and long-horizon program building.