Senior Hpc And Platform Engineer

About GeoSoftwareGeoSoftware is a global leader in Geoscience Software and provides a comprehensive range of software solutions that support our clients to achieve their goals. Our tools help our clients to develop a better understanding of the subsurface, manage exploration and development risks and quantify reserves more accurately with our advanced geoscience technology.
Working at GeoSoftware is all about passion and sharing our ambition for ongoing technological innovation. We have more than 150 employees, spread over 18 different countries. Our people are our most important asset and developing our talents is a priority. As part of GeoSoftware, you will work with international and diverse teams of experts in an environment of excellence which will enable you to create your own opportunities and realize your full potential.
Job SummaryWe are looking for an experienced HPC and platform engineer to join our Digital Platform team.
You would own the execution layer of our next generation of geoscience software: the layer that takes a piece of scientific work and runs it, at the same scale and with the same result, across a range of compute environments. That range covers a single workstation, on-premise HPC under Slurm or PBS, managed services such as Databricks, cloud batch on AWS and Azure, and bring-your-own-cloud deployment into an externally administered subscription. You would also own the build and delivery pipeline that gets it there.
Breadth is the point of the role. The same work has to run across all of those backends behind one consistent interface, deploy securely into environments we do not administer, and perform well against blob storage rather than local disk.
You would work in a small, senior, distributed team alongside computational algorithm and front-end engineers.
Duties and ResponsibilitiesDistributed Execution
Design and build the distributed execution framework: task sequencing and dependencies, elastic scaling, checkpoint and restart, retry with backoff, structured error reporting and preservation of partial output
Make the same workload run across local multiprocessing, Dask, Ray, Slurm and PBS, managed Spark services such as Databricks, and the managed cloud batch services on Azure, AWS and GCP — behind one consistent interface, with a clear view of what each backend is actually good for
Design job submission, status, progress and cost reporting as structured, introspectable data, so that work can be launched and monitored programmatically and automatically as well as by hand, with guards that prevent a mistyped job from running for hours
Build the job admission, entitlement and metering path
Cloud and Deployment
Build and maintain infrastructure as code — Terraform in particular — that stands the platform up repeatably in bring-your-own-cloud and externally administered environments
Containerize compute components as lean, independently versioned images
Implement the security surface of a deployed system: identity and access management, least‑privilege roles and service principals, secrets handling, network isolation, encryption in transit and at rest, and image and dependency scanning in the pipeline
Support restricted and disconnected deployment scenarios
Work with identity, blob and object storage, networking, quotas and cost control across at least one major cloud
Performance
Parallel decomposition and profiling of compute‑bound scientific workloads, so that throughput demonstrably scales with available CPU, GPU and storage bandwidth
Diagnose and fix I/O‑bound behaviour against blob and object storage: concurrency limits, throttling, request batching, and access patterns that are invisible on a local disk but cost minutes against a remote store
Establish and defend performance baselines, with measurement rather than assertion
Build, Release and Operations
Own CI/CD, build and release engineering: pipelines, a nightly regression harness, native installers for Windows and Linux, and containerized cloud delivery
Stand up observability: job lifecycle, resource utilization and cost visibility across every backend
Support what you build, and feed what you learn from doing so back into the design
Skills and CompetenciesSkills:
Production experience with at least one distributed compute framework—Ray, Dask, Spark or MPI—and a clear view of where each is the right tool. Having run one at real scale and watched it fail is worth more than having read about all of them
Strong Python for orchestration, services and bindings
Solid modern C++, or the confidence to work in it: the compute kernels are native code and the binding seam is shared with the algorithm engineers
Infrastructure as code, Terraform in particular, and experience deploying the same system into environments you do not control
Containerization and Kubernetes, including image layering and size discipline
Working proficiency in one or two major cloud providers—Azure, AWS or GCP—with awareness of the others
Blob and object storage performance in practice, not only in principle
Practical cloud security: identity and access management, least‑privilege roles, secrets management, network isolation and encryption
CI/CD pipelines, release engineering and the operational instincts that come from having supported what you built
Strong written English, and the habit of documenting asynchronously
Competencies:
Ownership and self-direction
Systems thinking across boundaries they do not personally own
Pragmatic judgement about where complexity is worth it
Collaboration across disciplines
Technical expertise (distributed computing and platform engineering)
Alignment with company interest
Qualifications & RequirementsRequired:
Bachelor's degree in computer science, engineering, physics, mathematics or a related technical discipline, or equivalent practical experience
Minimum 5 years of professional engineering experience in distributed computing, HPC or cloud platform engineering
Demonstrable production experience with a distributed compute framework
Demonstrable experience with infrastructure as code, and working proficiency in one or two major cloud providers
Experience owning a CI/CD and release pipeline end to end
Sound security practice in cloud and containerized deployment: identity and access management, secrets handling and least privilege
Able to work effectively in a small, remote, multi‑timezone team: writing things down, communicating asynchronously and operating without close supervision
Fluent written and spoken English
Desired:
HPC scheduler experience—Slurm or PBS—in a shared, contended environment
Managed cloud batch or Spark services: Databricks, Azure Batch or AWS Batch, and a view on when each beats a long‑lived cluster
Oil and gas or seismic processing background, or other large-scale scientific computing
Cloud‑native, chunked array storage formats and parallel I/O
OSDU, or other subsurface data platform integration
GPU computing: CUDA, or GPU‑accelerated numerical libraries
Security certification or formal security review experience;
supply‑chain security, SBOM or vulnerability management
Restricted or disconnected deployment;
software licensing and entitlement enforcement
Cost attribution and capping per job, per user and per tenant
Fluent with AI coding tools for prototyping, test generation and refactoring, with the judgement to know where their output must be verified rather than trusted
We value the unique ways in which we differ, and we are committed to equal employment opportunities for all professionals.
#J-18808-Ljbffr