WEKA Data Platform Administrator

Customer Success - Premium Services

About the job

WEKA is architecting a new approach to the enterprise data stack built for the age of reasoning. NeuralMesh by WEKA sets the standard for agentic AI data infrastructure with a cloud and AI-native software solution that can be deployed anywhere. It transforms legacy data silos into data pipelines that dramatically increase GPU utilization and make AI model training and inference, machine learning, and other compute-intensive workloads run faster, work more efficiently, and consume less energy.

WEKA is a pre-IPO, growth-stage company on a hyper-growth trajectory. We’ve raised $375M in capital with dozens of world-class venture capital and strategic investors. We help the world’s largest and most innovative enterprises and research organizations, including 12 of the Fortune 50, achieve discoveries, insights, and business outcomes faster and more sustainably. We’re passionate about solving our customers’ most complex data challenges to accelerate intelligent innovation and business value. If you share our passion, we invite you to join us on this exciting journey.

What You’ll Be Doing

Our AI Hyperscaler customers sell GPU capacity to other companies. Their tenants run training, inference, and checkpointing on WEKA. Those tenants expect the data layer to just work.

As a WEKA Data Platform Administrator, you own that outcome. You run the WEKA deployment for these cloud companies as an extension of their team, day to day, so they do not have to staff WEKA administration itself. You may work directly with their end customers to tune the infrastructure for the workloads they actually run.

Operate the Deployment

Run the WEKA environment day to day: monitoring, health, upgrades, capacity, configuration, and audits.
Create and manage file systems, tiering to object storage, snapshots, quotas, and protocol configuration (POSIX, NFS, SMB, S3) as tenants onboard and grow.
Own the operational run state. Keep the environment healthy, documented, and recoverable.
Execute approved changes to a standard: confirm scope and approval, classify risk, validate, keep a rollback, record the outcome.
Watch for drift, capacity pressure, and anomalies. Raise them early with a recommended path.

Optimize Infrastructure for NCP End Customers’ Workloads

May involve working directly with the customer’s tenants and their platform teams to understand how each workload uses storage: training, inference, checkpointing, data loading, and scheduling.
Tune WEKA to fit the workload. Filesystem layout, tiering, snapshot policy, quotas, and per-tenant QoS.
Validate and standardize the client layer: GPU node clients, BlueField-3 and SR-IOV networking, mount options, cgroups and kubelet CPU policy, and the WEKA Kubernetes operator and CSI driver.
Onboard new tenants and clients with a repeatable, verified procedure. Check configuration and scripts before they reach production, not during an escalation.
Establish performance baselines. Run benchmarks and tests with tenants, and report what changed and why.
Find scaling limits before customers do: client counts, process ceilings, NAT and routing behavior across tenant networks.

Runbooks, Automation, and Continuity

Build and maintain customer operating runbooks that a qualified backup can follow: procedures, access guidance, validation, rollback, and recovery steps.
Turn repeated manual work into scripts, automation, or checklists. Use Python, Bash, and Ansible.
Prepare handovers and backup briefs before time off or coverage changes.
Explain procedures and decisions to customer operators so they get more capable, not more dependent.

Escalation and Collaboration

Open and update cases with the observed condition, customer impact, work completed, evidence, and requested next step. Follow the established path to Technical Services, GRE, and R&D.
Keep the extended WEKA team informed of changes, risk, and anything the customer needs to hear.
Work with cloud providers, network teams, and other third parties in the customer’s environment to resolve issues.