S

SpAItial

AI for 3D environment generation

Machine Learning Systems & Infrastructure Engineer

Applied Machine LearningConfirmed live in the last 24 hours
Full-timeNo salary listedIn-person

About the Job

Requirements

  • At least 3 years of writing production-quality Python in a large, multi-author codebase; machine-learning systems experience is strongly preferred.
  • Hands-on experience with modern machine-learning training stacks such as PyTorch and DDP/FSDP or comparable technologies, including personally debugging distributed jobs across many GPUs and nodes.
  • Experience shipping non-trivial end-to-end data pipelines at scale, covering ingestion, transformation, validation, versioning, and republication; real-world sources with rate limits, authentication, or undocumented APIs are ideal.
  • Hands-on GPU compute and performance-debugging experience, including CUDA/NCCL, GPU utilization, networking bottlenecks, and profiling.
  • Working knowledge of cloud environments such as AWS, GCP, or Azure, including object storage, IAM, and cost awareness.
  • Proficiency with containers such as Docker and Kubernetes, and ability to read and write Terraform for the infrastructure surfaces shipped.
  • Strong working knowledge of storing and querying large datasets at scale, including SQL fundamentals; relational stores such as Postgres, analytical stores such as BigQuery or Snowflake, embedded stores such as SQLite, and object storage with caching layers.
  • Familiarity with ML workflow orchestration and experiment tracking, such as Kubeflow Pipelines and MLflow.
  • Experience with monitoring and observability tools such as Prometheus/Grafana and OpenTelemetry, and CI/CD for infrastructure and ML workflows such as GitHub Actions.

Responsibilities

  • Own and evolve the ML systems used for training, evaluation, and serving large foundation models, including trainer, dataset-loader, checkpointing, and experiment-orchestration code.
  • Improve high-throughput distributed-training stacks, such as PyTorch DDP/FSDP and NCCL, for performance, stability, and reproducibility, including preemption-safe and sharded checkpointing.
  • Build end-to-end Python pipelines that convert third-party capture sources into clean, versioned training datasets, including scraping and preprocessing.
  • Optimize storage at petabyte scale across object storage, FUSE mounts, caching layers, shared filesystems, and relational, analytical, or embedded metadata stores.
  • Operate systems researchers use to launch experiments, data jobs, and production endpoints, including workflow engines, GPU schedulers, experiment trackers, and managed-inference platforms; maintain a launcher SDK for one-command runs.
  • Ship workloads with Docker and Kubernetes; maintain Terraform infrastructure and CI/CD pipelines, including self-hosted GPU runners.
  • Implement monitoring, logging, and alerting for job performance, data-pipeline health, and cost; define SLOs and incident response for owned systems.
  • Manage secrets, IAM, and network boundaries, such as Tailscale and cloud VPC, for owned systems.
  • Partner with ML researchers, engineers, and the platform team to unblock training and data work and improve developer experience.

Desired Qualifications

  • Machine-learning systems experience is strongly preferred.
  • Real-world data sources involving rate limits, authentication, or undocumented APIs are ideal.

About the Company

SpAItial

AI for 3D environment generation

SpAItial builds Spatial Foundation Models that can understand, generate, and reason about 3D environments, including geometry, materials, and physics, starting from minimal inputs like a single image, a short video, or text. The product creates photorealistic, interactive virtual worlds beyond simple image generation, with applications in games, entertainment, CAD and construction visualization, VR/AR, and robotics simulation. The models run via licensing and APIs for developers, allowing integration into other software and services. SpAItial differentiates itself by focusing on 3D understanding and generation (not just 2D images) and by combining perception, generation, and physical reasoning to produce usable 3D scenes. The company’s goal is to enable developers and creators to build realistic, interactive 3D environments quickly and at scale, enabling a broad set of industries to adopt advanced AI-driven 3D tools.

Company size
11-50
Company stage
Seed
Total funding
$13M
Headquarters
London, United Kingdom
Founded
2024

Recently Posted Jobs

Explore more opportunities at SpAItial

See more jobs