
NexusAI architecture
Six layers on one Kubernetes stack, from deployment and data to reasoning and solutions, with AI Ops across every layer.
What is the NexusAI architecture?
NexusAI is built as six layers on one Kubernetes stack. A data plane reads your systems in place, a control plane holds the ontology, policies and workflows, an AI engine runs AI jobs on distributed compute, a reasoning plane lets agents plan and act through approved tools, and a solutions layer packages it all for each industry. AI Ops checks the health of every layer.
It runs on your servers, in a sovereign cloud or hybrid, so your data stays where it is allowed to live.
Data plane
Structured, unstructured, streaming, graph and vector data, read in place and unified for reasoning.
Control plane
Schemas, policy guardrails, workflows and integration rules that keep every action governed and auditable.
AI engine and reasoning plane
AI jobs on distributed compute; agents that plan, call defined tools and act within policy.
Solutions layer
Packaged applications and copilots for telecom, supply chain, finance, revenue operations, healthcare and IT operations.
Processing
& Cache
Compute
Telemetry
Data
Data
Data
Data
Data
Events
Hover a layer to highlight it
One stack, layer by layer.
Solutions call the reasoning plane; the AI engine runs jobs on distributed compute; the control plane governs every action; the data plane unifies every kind of data; and it all deploys wherever your data must stay.
AI-native apps and identity services on one platform. Packaged solutions and copilots your teams use every day.
Agents plan, call tools and act. ReAct planning turns a goal into a sequence of governed actions.
Executes AI jobs on distributed computing — from data prep and fine-tuning to real-time inference.
Schemas, policy guardrails, workflows and integration rules keep every action governed and auditable.
Structured, unstructured, streaming, graph and vector data unified in one place, ready for reasoning.
Runs where your data must stay — cloud, on-premise, hybrid or edge — as one containerised stack.

Where does NexusAI run?
Wherever your data must stay. NexusAI ships as one Kubernetes-based stack that installs with a single operator in your data centre, in a sovereign cloud, at the edge, or across all three.
- One stack on Kubernetes, with Spark and Ray for distributed compute
- CPU, GPU and TPU execution for training, fine-tuning and real-time inference
- Policy guardrails, observability and security in every layer
- Connectors, ETL and change data capture, with nothing to rip out
Questions leaders ask
What are the layers of the NexusAI platform?
Six layers on one stack: deployment and delivery, the data plane, the control plane, the AI engine, the reasoning plane and the solutions layer, with AI Ops (diagnostics and health) running across all of them.
Where does NexusAI run?
On your own servers, in a sovereign cloud, or in a hybrid set-up. It installs with one Kubernetes operator, so regulated data does not have to leave your environment.
Does NexusAI copy our data into a new database?
No. The data plane reads your systems where they live through a virtual data layer, and mapping rules turn tables into business objects. Queries go to the source.
How does NexusAI run heavy AI workloads?
The AI engine runs AI jobs on distributed compute, with Kubernetes, Spark and Ray, on CPUs, GPUs or TPUs. A worker pool scales with the job queue, and monitoring shows what each job used.
How are AI actions governed?
The control plane holds schemas, policy guardrails, workflows and integration rules. Policies sit outside the model, high-risk actions wait for a person, and every action is logged.
Get started
See the architecture on your data
Turn enterprise systems into an intelligent operating layer that understands your business and executes operations autonomously.