- The deployment, from the first architecture conversation through go-live and into steady-state operation.
- Its integration with the customer's identity, data and engineering systems, and the infrastructure and model serving layer beneath it.
- The relationship with their IT, security and platform teams: assessments, architecture reviews, and the call when something breaks.
- The diagnosis when it does break, including when the cause sits outside our own software.
- The runbooks, post-mortems and playbooks that make the next deployment easier, and the feedback loop into Product.
Where you will do it Managed Kubernetes across the major clouds. On-premises and fully air-gapped clusters with no route to the internet, GPU infrastructure, self-hosted model serving. Customer HPC environments, schedulers, and shared storage under real load.
If installing software with no internet access, on hardware you do not control, reads as the interesting part of the job rather than the obstacle, this will suit you.