The problem
Spatial data has never been more abundant, or more fragmented. The systems that hold it do not talk to each other.
Different worlds. Same question. Where?
Spatial intelligence for decisions that depend on where.
A supply chain on Earth and a landing site on the Moon look like different worlds. They begin with the same question: where is the risk, where should we act, where is it safe, where could we stay. Anassa.ai is an independent research lab studying how autonomous AI agents can turn that question into decisions, at machine speed and under human governance. What if spatial decisions took seconds, not hours?
Research
One question. Many places.
The gap
People end up as the connective tissue, moving and reconciling data by hand while the question that matters waits.
The research
What changes when governed agents do that coordination instead, and how we would know.
Where should we build, protect, or intervene?
Cross-source data reconciliation
- For example
- Two agencies' parcel boundaries disagree by metres along a floodplain.
- What we would measure
- Pipelines that complete without a manual restart; reconciled records checked against the authoritative source.
Where is disruption emerging?
Supply chain risk
- For example
- A warehouse cluster sits inside a flood zone as the monsoon season starts.
- What we would measure
- Minutes from a hazard update to a refreshed risk map; analyst hours per assessment.
Where is it safe to move and act?
Autonomous vehicle safety
- For example
- An unsignalised intersection with ten times the incident rate of the surrounding highway.
- What we would measure
- Simulation coverage gained per unit of compute.
Where could humanity establish a lasting presence?
Lunar site selection
- For example
- A south-pole crater rim balancing sunlight, water ice and line-of-sight to Earth.
- What we would measure
- Specialist review time; every numerical claim traced to a primary source.
Framework
ANASSA
Autonomous Neural Agents for Spatial Systems Architecture
In plain terms: an air-traffic controller for spatial data. It does not fly the planes; it decides which agent does what, in what order, and stops anything that cannot be verified.
The ANASSA architecture defines an orchestrator for coordinating 47+ specialized agent roles, organized into eleven components across four layers and a continuous cognitive loop. The framework is being developed to validate spatial and numerical outputs against authoritative ground truth before they influence downstream decisions, with Human-in-the-Loop governance at key decision points.
- Sits above
- Existing geospatial platforms, data pipelines and compute. Nothing is replaced.
- Connects through
- MCP, A2A, Geospatial SDKs and APIs, IoT connectors, legacy system adapters, and others.
- Ground truth
- Certified Earth and Space data in versioned repositories with provenance and trust scores.
- Read more
- White paper (PDF)
- 01Sense
- 02Reason
- 03PlanHuman review gate
- 04DecideHuman review gate
- 05Execute
- 06Learn
Where is a first-class input.
Location, time and context enter reasoning at the start, not as a coordinate attached at the end.
Humans govern. Agents act.
Review gates at planning and decision keep the highest-stakes choices under explicit accountability.
Trust is verified.
ANASSA is designed to check spatial and numerical claims against authoritative ground truth before they can influence a decision.
Above, not instead of.
The orchestration layer sits on top of existing geospatial platforms, data and compute rather than replacing them.
Research collaboration
Three ways to collaborate.
Industry
Research exchange and use-case exploration with organizations interested in Spatial AI, agent orchestration, geospatial intelligence, governance, and related research questions.
Researchers
Joint studies, shared benchmarks, replication, and co-authored research on agent orchestration, spatial reasoning, Human-in-the-Loop governance, and decentralized intelligence.
Academia
Architecture, governance models, research questions, and use cases documented for citation, teaching, experimentation, and replication. Departments, labs, faculty, and graduate researchers are welcome.
About
An independent lab, asking one question across two worlds.
Anassa.ai is an independent Spatial Intelligence research lab; ANASSA is the research framework being developed and evaluated through that work. The purpose is to help people and intelligent systems make better decisions about where.
- Founded
- 2025
- Focus
- Agentic orchestration for spatial decisions
- Contact
- contact@anassa.ai
The story behind the lab
Before the lab, its founder spent years as a technical program manager, the person whose job was to hold the systems together. That experience is where ANASSA comes from.
"The honest description of the job was being the glue. Engineers built in one tool, work was tracked in another, and alignment happened in weekly syncs that existed only because the tools didn't talk. I came to call it the fragmentation tax: at scale, 30 to 40 percent of a team's capacity spent holding the system together instead of moving it forward.
When agentic systems arrived, I argued that the role wasn't disappearing, it was moving upstream: from coordinating people and tracking status to writing intent clearly enough for agents to act on it, setting the guardrails, and making the calls that still need a human.
Then I looked at spatial decisions, the domain of my research, and saw the same tax at far higher stakes. Analysts reconciling geospatial platforms, satellite feeds and legacy systems by hand, while the question that matters, where, waits. A vague goal there doesn't just slow things down. It gets executed confidently in the wrong place.
Anassa.ai is that observation turned into an applied research program. Agents do the coordination at machine speed. People hold the intent, the constraints and the decision gates. I've been the glue. This is what I think replaces it."
Publications
Latest work.
Published research
ANASSA: An Agentic AI Orchestration Framework for Spatial Intelligence
Constantinos Papantoniou · Brian Hilton
Defines the ANASSA architecture: eleven components across four layers, a six-step cognitive loop, agent orchestration, provenance, validation, governance, and human decision authority.
ANASSA: Autonomous Neural Agents for Spatial Systems Architecture
Anassa.ai
Foundational white paper describing the ANASSA architecture, orchestration model, governance mechanisms, and Spatial Intelligence research direction.
Research in progress
Agentic AI Orchestration Framework for Spatial Intelligence: Supply Chain Deployment
Constantinos Papantoniou · Brian Hilton
Research artifacts
ANASSA high-level architecture
Anassa.ai · v1.0
ANASSA low-level architecture
Anassa.ai · v1.0
Different worlds. Same question. Where?
If your research, domain, or technical challenge depends on where, we welcome conversations about research collaboration, academic exchange, and future areas of study.