Field EngineeringRemote, with client-site travelRef: DL-2026-006
Forward Deployed Engineer, Inference Infrastructure
Stand up the platform inside a client's own Kubernetes cluster or air-gapped network, and own the inference layer it runs on once it's there.
What you'd do
- Deploy the platform directly into a client's environment under our BYOC, private cloud and embed models: their VPC, their Kubernetes cluster, their IAM and KMS, sometimes air-gapped with no route out at all.
- Own inference infrastructure end to end: model serving, GPU scheduling and autoscaling, and the routing layer that lets a deployment run on a frontier API, open weights on the client's own hardware, or a dedicated inference endpoint, under the client's own key, without the application layer caring which.
- Sit embedded with a client's platform or infra team for the length of a deployment, translating "it works in our environment" into "it works inside your constraints": their network policy, their change windows, their existing observability stack.
- Write and maintain the infrastructure as code (Terraform across cloud providers, Docker Compose or Kubernetes manifests for the embed model) that makes a deployment repeatable instead of a one-off script someone has to remember.
- Own the SLOs for whatever you deployed after you deploy it: latency, cost per request, and uptime, with the pager to match.
What we're looking for
- Production experience running Kubernetes yourself, not just deploying to a cluster someone else built and tuned: scheduling, resource limits, network policy, and what actually happens when a node goes down mid-request.
- Hands-on experience serving models in production (vLLM, Triton, TGI, or equivalent), including the GPU scheduling and batching tradeoffs that decide whether a deployment meets its latency target under real load.
- Comfortable being the engineer on-site or on a call with a client's infra team, translating between what we built and what their environment actually allows.
- Solid infrastructure-as-code fundamentals across at least one major cloud plus a credible story for the air-gapped or fully on-prem case, where the usual managed services are not available.
- Willingness to travel to a client site when a deployment genuinely needs a person in the room, not just a screen share.
Ready to apply?
Fill in the form on the next page, CV included; one submission sends it straight to us.