JAX
| ONE PLATFORM · EVERY SPATIAL DOMAIN |
One platform.
Every spatial domain.
Jax is Nexma's spatial intelligence agent. It reads and writes the DataStore, reasons against the active Ontology, runs the Math Engine, and produces the artifact you'd otherwise build by hand.
AGENT INDEX · 02
one platform,every spatialdomain.
The work used to live across a stack — a spreadsheet to scope it, a GIS desktop to place it, a CAD tool to draft it, a scheduler to dispatch it, a chat channel to chase it. Five surfaces, five sources of truth, no audit trail that survives the handoff. Jax collapses the stack into one loop. The DataStore holds the world, the Ontology types it, the Math Engine solves it, and the agent runs the cycle — engineering one minute, investigation the next, briefings on demand.
And this is what that changes:
The stack collapses into one loop.
Eight surfaces of the same agent.
- 02AAGENT
Autonomous Engineering
Fiber and utility networks planned and designed end to end.
Explore - 02BAGENT
Investigation
OSINT, threat geolocation, and cyber investigation as one loop.
Explore - 02CAGENT
Object Detection
State-of-the-art object detection in images and video.
Explore - 02DAGENT
Satellite Imagery
Agentic access to satellite imagery and geospatial analytics.
Explore - 02EAGENT
Operations Management
Track assets, dispatch crews, manage tasks, automate workflows.
Explore - 02FAGENT
Real-Time Digital Twin
Your operations mirrored as a live, governed graph.
Explore - 02GAGENT
Spatial Analysis
Spatial queries, network and terrain analysis at machine speed.
Explore - 02HAGENT
Slides & Spreadsheets
Decision-ready briefings and structured analyses on demand.
Explore
The same loop, every time.
Whether the question is “route a fiber feeder” or “summarize last week's threat picture”, Jax runs the same five-step cycle — bound to the same DataStore, governed by the same Ontology, solved by the same Math Engine. The priors change, the loop does not.
- 01READ
Pull the relevant slice of the DataStore — entities, geometry, history, feeds.
- 02REASON
Bind the question to the active Ontology. Decide which constraints and skills apply.
- 03SOLVE
Compose the eight primitives, dispatch the Math Engine, iterate to a feasible answer.
- 04WRITE
Commit the result back to the DataStore with provenance — every artifact audit-trailed.
- 05RENDER
Surface to whichever view the human is in: globe, tab, graph, briefing.
Why one platform
and not eight.
- 01ONE GOVERNANCE BOUNDARY
Eight agents means eight permission models, eight failure modes, eight things to sign off. One platform means one trust contract — what the operator approves, the agent does, on the record.
- 02ONE AUDIT TRAIL
Every DataStore write — engineering, investigation, briefing — lands in the same log. The provenance graph survives the surface change. Auditors get a single source, not a stitch.
- 03ONE ONTOLOGY BINDING
The world is typed once. Skills inherit. A field crew's location is the same entity the planner reasons about and the briefing cites — not three reconciled copies in three tools.
- 04ONE MATH ENGINE
MIP, VRP, CP, simulation, and heuristic dispatch live behind one solve primitive. Every capability shares the optimization investment instead of buying it five times over.
- 05SWAPPABLE SKILLS
Domains plug in as skills, not as new agents. Add water, retire FTTH, swap defense for logistics — the foundation is unchanged. The org learns the agent once.