Case Study: Modernising and automating Defence simulation infrastructure with AI-driven orchestration
Client: Department of Defence, Joint Collective Training Branch
Project: Defence Training and Experimentation Network (DTEN)
How Profectus transformed a fragmented, manually built Defence simulation environment into a modular, automated and continuously monitored platform, adding a locally hosted LLM and RAG orchestration tool that turns natural language exercise descriptions into infrastructure builds.
Client and context
The Joint Collective Training Branch is responsible for designing and delivering joint collective training for the Australian Defence Force, using the Defence Training and Experimentation Network (DTEN) as its simulation environment. DTEN is a complex, federated environment comprising live, virtual and constructive components integrated across compute platforms, simulation hosts, network infrastructure, middleware, data services and specialist peripherals, deployed across classified and unclassified domains.
The digital infrastructure supporting these high-fidelity simulation exercises relied on a range of legacy systems and manual processes that limited scalability, slowed delivery and constrained operational readiness. As training tempo increased, this model was no longer sustainable, and Defence required an environment that could be rapidly deployed, securely managed and continuously improved.
The engagement
Profectus was engaged to modernise and sustain the infrastructure underpinning DTEN, transforming it end to end into a modular, automated environment that could be rapidly deployed and continuously improved. The remit spanned rebuilding simulation software into modular packages, establishing CI/CD automation, virtualising legacy platforms, standing up integrated monitoring, and prototyping a locally hosted, AI-driven orchestration capability, all delivered within disciplined engineering and sustainment governance aligned to Defence operating constraints.
The challenge
The existing environment placed several compounding demands on delivery:
- Fragmented estate: a bespoke mix of applications, static infrastructure and undocumented legacy components, with system knowledge held informally across disparate sources
- Manual, error-prone builds: each simulation instance required repeated setup from scratch and specialised knowledge, limiting scalability and consistency
- No integrated visibility: limited insight into system health, with no integrated monitoring across environments to support proactive management
- Requirements to build gap: no capacity to translate high-level operational requirements into executable technical builds without heavy human input
Our approach
Profectus applied a systems engineering approach to modernise the environment across four coordinated workstreams.
Modular packaging and CI/CD
The team rebuilt key simulation software into modular Red Hat RPM packages, established CI/CD automation with Ansible Tower and Git, and created tailored standard operating environments for RHEL 7, 8 and 9. Custom pipelines automated environment creation from scenario inputs, reducing manual effort and improving consistency and repeatability.
Virtualisation and monitoring
Infrastructure was migrated from physical to virtualised platforms using VMware ESXi, VSCA and VSAN across multi-node, multi-domain deployments. Integrated monitoring and alerting stacks using Grafana and Prometheus provided real-time insight and historic trend data, giving full-stack visibility across every environment.
AI-driven orchestration
Profectus developed a locally hosted LLM-driven orchestration tool that converts user-submitted exercise descriptions into build proposals by referencing internal documentation via retrieval augmented generation (RAG). Once approved, the system triggers the infrastructure pipeline, effectively enabling natural language driven environment builds. This capability was tested and integrated incrementally.
Federation engineering and modelling
Across DTEN, Profectus aligned federation and environment engineering to the Distributed Simulation Engineering and Execution Process (DSEEP) and supported Joint Live Virtual Constructive (JLVC) style environments. The team built a structured, maintainable model of the environment, mapped dependencies and configuration states, and aligned modelling to the Joint Event Life Cycle (JELC) so environments could be reliably prepared, executed, reset and sustained.
What we delivered
- Rapid deployment: automated build processes reduced environment deployment time from days to under an hour
- Sustained exercise support: a range of simulation exercises, from small-scale rehearsals to full-spectrum operational scenarios, supported on the new infrastructure over a 12-month period across classified and unclassified domains
- Full-stack visibility: dashboards providing complete visibility across all environments, improving proactive management and reducing unplanned outages
- Legacy modernisation: legacy applications now operating reliably in virtualised environments with full integration
- Operational AI prototype: a locally hosted AI orchestration tool operating in prototype form, accurately generating infrastructure configurations from user input and reducing planning effort
Systems and methods
Red Hat RPM · Ansible Tower · Git · VMware ESXi, VSCA and VSAN · Grafana and Prometheus · Locally hosted LLM with RAG · DSEEP · JLVC · JELC
The result
This unified platform has positioned Defence for future readiness, enabling greater exercise frequency, improved technical reliability and faster turnaround between scenarios. By automating infrastructure and enabling AI-assisted orchestration, Profectus reduced technical overhead while empowering non-technical users to drive outcomes, and combined disciplined, DSEEP-aligned federation engineering with repeatable automated deployment to sustain reliable, repeatable training on DTEN. The solution remains in active use, scalable to future needs and adaptable to new technologies.