ML Infra Engineer
London (hybrid)
£100-150k base + equity
Built a GPU training stack from scratch before?
Designed large-scale data pipelines end-to-end?
Happy owning infra + ML systems in a fast, early startup?
We’re an early-stage generative AI company developing state-of-the-art diffusion transformer models for 3D generation. We're training from scratch, at scale. Backed by top-tier investors and fresh off a $7M seed, we’re now hiring an ML Infrastructure Engineer to build our data and training stack from the ground up.
Our customers include global companies in autonomous driving and robotics.
The role
* Design and own the end-to-end data pipeline powering large-scale 3D pretraining: ingestion, storage, preprocessing, and streaming.
* Implement infrastructure-as-code, observability, CI/CD, and scalable deployment practices.
* Enable fast experimentation for diffusion transformer training (PyTorch/JAX, DDP, mixed precision, distributed compute).
* Design and scale our GPU training cluster
* Work directly with the founding team to set technical direction and engineering culture.
You
* Strong experience in ML platform / ML infra / MLOps roles, ideally at an AI or high-performance compute company.
* Deep familiarity with GPU orchestration (K8s + NVIDIA stack, Slurm, Ray, etc.).
* Comfort standing up cloud infrastructure from scratch (AWS preferred).
* Experience building data pipelines (Airflow/Dagster, Spark/Beam, Kafka, Parquet/S3).
* A bias toward ownership, pragmatism, and building quickly with high quality.
* London-based or willing to relocate — onsite 3 days per week.
Why Join
* Build the core infrastructure behind frontier diffusion transformers for 3D.
* Join at the earliest stage — massive technical ownership, fast execution, zero legacy constraints.
* Competitive compensation + meaningful equity.
* Work closely with founders and researchers shipping cutting-edge generative models.
* A chance to define a new category in generative 3D.