Compute
A dedicated training fleet
The cluster behind NEXUS Core runs at sustained high utilisation, driving the cost per training token down against the original plan.
Participant portal · Programme overview
Mintelligent Services
Seven AI modules.
One converged platform.
We are building an applied intelligence platform end to end — the data refinery, the reasoning model, perception, autonomous agents, the safety control plane, the deployment fabric and the enterprise product surface. Each part is a module that stands alone. Together they become one system.
Slide 1 of 8: The platform
The platform
Seven modules converging into one applied intelligence system.
How it works
From raw data to a working system
The seven modules are a pipeline, not a portfolio. Each one hands its output to the next, and the last one turns the whole stack into a product. Step through the pipeline below.
Think of Mintelligent Services as building one intelligent factory, made of seven specialised machines that each do one job and pass their work to the next. The goal is to turn the messy information that businesses already have — documents, images, spreadsheets, sensor feeds, recordings — into reliable, traceable AI work that regulated industries can actually use.
SENTRA cleans and traces the raw material. NEXUS Core is the brain that learns to reason from it. ORACLE gives that brain eyes and ears, so it can read photos, scans and video. HELIX turns answers into completed tasks by operating real business tools. AEGIS runs the safety and audit checks. VERTEX delivers the whole system wherever the customer needs it, and ATLAS is the product surface that people actually log into.
Each stage means something concrete: SENTRA makes the data trustworthy, NEXUS makes the model smart, ORACLE makes it multi-modal, HELIX makes it useful, AEGIS makes it safe, VERTEX makes it deployable, and ATLAS makes it sellable. Together they form a single platform that can be inspected, governed and deployed in places where AI has to prove itself before it is trusted.
Step 01 · SENTRA
Data is refined and traced
SENTRA is the refinery at the front of the platform. It ingests licensed and permissive sources, deduplicates them at scale, scores every document for quality and toxicity, and attaches a signed provenance record before anything is allowed near a training run.
- Near-duplicate detection across billions of documents
- Quality, safety and licence scoring per document
- Signed provenance chain that survives into model releases
Takes in
Licensed corpora, customer archives, public permissive data
Hands on
Curated, deduplicated, provenance-stamped training shards
Participation cap
US$14,300,000 across seven modules
The programme is capped at A$22,000,000, converted at A$1.00 = US$0.65. Select a module to see how much of the cap it carries — or click through to its full detail page.
M0123.8%of the cap
Cap
US$14,300,000
Allocated
US$14,300,000
Unallocated
US$0
NEXUS Core — Foundation model and inference engine
Open NEXUS CoreInside the programme
Where the work actually happens
Compute, research and deployment run in parallel across the seven modules.
Research
Evaluation before promotion
Every checkpoint clears blocking evaluations and adversarial suites before it is allowed to move downstream to the other modules.
Deployment
Built to run in production
VERTEX and ATLAS take the platform into customer environments, from datacentre fleets to constrained on-premise hardware.
Module register
What each module does
A plain-English summary of each part of the platform — what the software actually does, in one line. Open a module for the full explanation, technical detail and capital allocation.
M01
NEXUS Core
Foundation model and inference engine
The reasoning substrate every other module depends on: a 70B-parameter mixture-of-experts foundation model with a custom low-latency inference runtime.
In plain English
The brain of the platform. It reads a question, works through it step by step, and writes back an answer — fast enough and cheap enough to run all day inside a business.
62% completeQ3 2025 — Q2 2027
M02
SENTRA
Data ingestion and synthetic corpus pipeline
The industrial-scale data refinery that feeds NEXUS: licensed acquisition, provenance tracking, deduplication, and governed synthetic data generation.
In plain English
The refinery that turns messy, mixed-up source material into clean training data — and keeps a receipt showing where every document came from.
74% completeQ2 2025 — Q4 2026
M03
ORACLE
Multimodal reasoning and vision stack
Vision, document and sensor understanding fused into the core reasoning loop, so the platform can act on the physical and paper world, not just text.
In plain English
Gives the platform eyes and ears. It reads scanned paperwork, photographs, diagrams, charts and audio, and hands the meaning to the reasoning model.
41% completeQ4 2025 — Q3 2027
M04
HELIX
Autonomous agent orchestration layer
Long-horizon agents that plan, call tools, recover from failure, and hand work between each other with a durable execution record.
In plain English
Turns answers into completed work. Agents plan a task, use real tools and systems to carry it out, and stop for human approval where the stakes are high.
35% completeQ1 2026 — Q4 2027
M05
AEGIS
Safety, alignment and compliance framework
The control plane that makes the platform sellable into regulated markets: red-teaming, guardrails, evidence generation and jurisdictional policy packs.
In plain English
The control plane and the brakes. It tests every release for harmful, biased or leaky behaviour, enforces policy live, and keeps an audit trail regulators can read.
48% completeQ3 2025 — Q2 2027
M06
VERTEX
Edge deployment and compute fabric
The distribution layer: distilled models and a scheduling fabric that run the platform on-premise, air-gapped, or at the edge with the same API.
In plain English
The plumbing that makes the platform actually run — scheduling scarce hardware efficiently and shrinking models so they fit on modest machines.
29% completeQ2 2026 — Q1 2028
M07
ATLAS
Enterprise interface and integration suite
The commercial surface customers actually touch: workspace, admin console, connectors, billing and the public API that turns the stack into revenue.
In plain English
The part customers actually see and buy — the workspace, the connectors into their existing systems, and the controls their administrators need.
22% completeQ3 2026 — Q2 2028
Signal
Most recent programme updates
Milestone2026-07-18
NEXUS 13B pilot clears internal reasoning gate
The pilot checkpoint passed all twelve blocking evaluations, including the multi-step tool-use suite that failed in April. Promotion to the 70B production run is approved for August.
Financial2026-06-30
H1 FY26 capital deployment report issued
38% of committed capital deployed against a 41% plan. Underspend is concentrated in VERTEX hardware validation, which shifts to H2 following a revised partner schedule.
Commercial2026-06-11
Two additional design partners signed for ATLAS
A logistics group and a regional utility joined the ATLAS design partner programme, taking the cohort to five and covering three regulated verticals.