EC2-compatible virtual machine orchestration built on QEMU. Public sector teams launch and manage instances using the AWS CLI, SDKs, and Terraform they already use, with cloud-init metadata, instance type definitions, and full lifecycle management — all running on operator-owned hardware with zero cloud dependency.
EBS-compatible persistent block storage with write-ahead-log durability, snapshots, replication, and NVMe caching. Volumes attach via NBD or virtio-blk, giving mission databases, geospatial systems, and AI/ML workloads cloud-equivalent storage performance at the edge.
S3-compatible object storage with Signature V4 authentication, multipart uploads, and Reed-Solomon erasure coding for multi-node fault tolerance. Designed to keep imagery, model artifacts, logs, and mission data available even when individual nodes are lost.
A pre-integrated runtime that lets agencies move existing AWS GovCloud workloads onto sovereign, operator-controlled hardware without re-architecting. Targeted at programs subject to data residency, classification, or DDIL requirements that hyperscale cloud cannot satisfy.
Pre-validated edge appliances for trials, exercises, and rapid fielding. Customers can also bring their own hardware from Mulga's Validated Hardware List of tested platforms — Spinifex is hardware-agnostic by design and is being delivered alongside reference architectures with leading OEM partners.
Run inference, retraining, and model-serving workloads on operator hardware with no reach-back to a hyperscaler. Models field, swap, or roll back as standard mission workloads without re-integration or touching the underlying operating system.
A common execution layer that mission platforms — command and control systems, autonomy stacks, common operating pictures, collaboration tools — run on top of. Mulga handles execution survivability so vendors and integrators focus on mission logic, not infrastructure plumbing.
Aligns with distributed and coalition architectures (including JADC2-style frameworks) by standardizing execution without forcing convergence on a single cloud provider or vendor stack. Allies and partners run interoperable workloads on locally authoritative nodes.