Spatial AI Research

Understanding real
environments over time

Mihawk turns the physical infrastructure you already have into a persistent world model — one that remembers how spaces, people, vehicles, and events evolve, and can be queried, reasoned over, simulated, and acted on.

The problem

Vision sees. Intelligence remembers.

Today's systems do one — not both.

Traditional — stateless
Mihawk — persistent

Alerts need speed; identity needs time — real understanding needs both.

Per-frame accuracy doesn't produce understanding.

Our approach

Perception, memory, and reasoning — composed over time

Mihawk maintains a structured, time-indexed record of what actually happened — queryable, and constrained by recorded evidence.

01
Perceive

Structure from imperfect observations

Observations from existing sensors are fused into a shared, persistent model of the environment. Positions, relationships, and events are grounded in space and time — so the system knows where things are, even across viewpoints, occlusions, and changing layouts.

Heterogeneous observation3D scene structureCross-view alignment
02
Remember

Memory over time

People and objects are tracked across views and over time into a searchable record of who was where, when.

Entity continuityTrajectory memoryEvent history
03
Reason

Inference under constraints

Reasoning agents build and test hypotheses over the model, ruling out what physical and spatial constraints make impossible.

Hypothesis assemblyScenario replayCounterfactual check
W[t₀, t₁] = { Entities, Geometry, Relations, Time }
Design philosophy

Structure over precision

“To reason about the physical world at scale, a system must not only capture structure — it must think with it.”

Relational first

Mihawk understands a space through how things relate and persist — not through perfect measurement — so it holds up on real, imperfect deployments.

Built for infrastructure that changes

Cameras move, sensors are added, layouts change, and real environments rarely stay clean. Mihawk folds new and imperfect signals into the same persistent world model without rebuilding the deployment from scratch.

Sufficient fidelity, not perfect simulation

Not a perfect simulator — just enough structure to constrain what's possible and compare candidate futures, degrading gracefully rather than collapsing.

Research

A research-product company

The principles behind the product are formalized and benchmarked in our research — not asserted.

PHAST: Grey-Box Port-Hamiltonian State-Space Models with Recoverable Physical Structure

Structure-preserving dynamics — up to ~10× smaller and ~10⁴× more accurate than Transformers across 13 SciML benchmarks.

Paper

C-PHAST: Compositional Port-Hamiltonian World Models for Structured Dynamics Transfer

A physical critic that flags unsafe actions before failure — one structured world model across robotics and DC microgrids: ~10× more stable under a robot-embodiment swap (graph baselines diverge), robust under topology and sensing change.

Preprint
Systems in deployment

One world model, many physical problems

Mihawk runs today on infrastructure already in place — under real constraints, across different environments.

Mihawk adds intelligence before it adds hardware.

It works with the cameras and systems you already have, and keeps working as those environments change.

Live deployment
Education

Attendance

From a single existing classroom camera, Mihawk maintains a persistent record of who was present and when — despite occlusion, natural movement, and imperfect camera placement.

Live deployment
Security

Access control

Entry and exit decisions resolved with identity, policy, and event context on top of existing access infrastructure.

Live deployment
Traffic

Vehicle violations

Vehicle movement, violations, and evidence reconstructed from existing roadside cameras — the same world model applied to vehicles.

Each deployment is a different query over the same underlying world model. Persistent situational understanding is the capability.

Understanding real environments over time

If you operate in environments where context, accountability, and time matter — or if you are building in physical-world intelligence — we'd like to hear from you.