Expanse raises seed round as GPU workloads strain cluster efficiency
Expanse predicts compute, memory and runtime needs before jobs reach a scheduler, helping AI and research teams reduce failed runs and idle capacity inside their own infrastructure.
Right-size a job before it runs
Expanse connects to an existing SLURM, Kubernetes or Nomad cluster and learns from workload history. Before a job enters the queue, it estimates GPU, memory and runtime needs and shows the recommendation with confidence and supporting evidence. Teams can adjust requests before an undersized job fails or an oversized allocation reserves capacity it will not use.
The feedback loop continues after execution: the system can diagnose failures from code, logs and metrics, then use observed outcomes to improve future predictions. Expanse’s models and management layer run within customer-controlled infrastructure, keeping code and telemetry on the cluster.
Turning infrastructure waste into a planning signal
The company says one production deployment identified idle compute capacity within a month. That illustrates the operational value of treating job sizing as a prediction problem: infrastructure teams can make capacity visible before workloads consume it, while keeping their existing cluster and submission workflow in place.
The seed financing backs a product already being deployed and gives Expanse room to broaden adoption among organisations running AI, high-performance computing and research workloads.
Use of funds
Expand the engineering team
Accelerate product development
Reach wider industry adoption