Jobs at Diffractive Labs

Build and own cloud infrastructure and GPU compute environments for ML model training and deployment. Implement IaC, CI/CD, orchestration, containerisation, monitoring, security, and internal tooling to enable reproducible, scalable research-to-production ML workflows.

Develop, train, and fine-tune machine-learned interatomic potentials (MLIPs) for magnetic and structural materials. Design DFT dataset workflows, active learning loops, and extend MLIP architectures to capture spin-lattice interactions. Build reproducible model and dataset pipelines and collaborate with materials scientists to translate physical intuition into training objectives.

Operate and maintain research-grade synthesis and characterisation instruments (XRD, SEM, vacuum systems, high-field magnets, high-temperature metallurgy). Run reproducible experiments, perform preventive maintenance and fault diagnosis, write SOPs and risk assessments, manage consumables and inventory, collaborate with materials scientists and ML researchers, and support integration of experimental data into digital/AI pipelines.

Implement and scale cutting-edge ML research into production-ready systems: build distributed training and inference on GPU clusters, optimize model performance, create evaluation and experiment-tracking tooling, architect multimodal scientific data pipelines, and collaborate with researchers and materials scientists to deploy models in closed-loop hardware environments.

Build and maintain software and orchestration that turns a physical lab into an autonomous system: implement instrument drivers, coordinate robots and data pipelines, automate workflows, integrate new instruments, and ensure reliability, error handling, and reproducibility while collaborating with materials scientists and ML researchers.

Design and own data architecture and ingestion pipelines that convert wet-lab instrument outputs into versioned, reproducible datasets for ML and RL training. Build tooling for dataset inspection, lineage, and integration with model training workflows, and collaborate closely with ML researchers to support pretraining, midtraining, and RL experiments.

Develop and implement advanced theoretical and computational methods beyond standard DFT to predict intrinsic magnetic properties. Apply GW/DMFT and Green's-function many-body techniques, derive analytic formalisms, implement numerically, integrate with DFT production workflows, and collaborate with ML teams to improve predictive models and training data.

Lead ML research for an AI-driven materials discovery platform: design and prototype architectures (GNNs, generative, foundation models), develop active learning/optimization strategies, address representation learning for sparse noisy experimental data, collaborate with materials scientists to encode physics into models, and drive proofs-of-concept to production with engineering partners.