Research Engineer, Multimodal Brain Signal AI

Sigma Nova; Paris; Hybrid; EUR 70,000 - 120,000; Full time

Job summary

Sigma Nova is building foundational models for scientific signals (EEG, fMRI, ultrasound) and aims to extend to other domains. The architect of our Data & Training Factory will scale pre-training, fine-tune models, and engineer robust multimodal pipelines with a focus on data integrity and reproducibility.

The role requires 3–7 years in applied research, deep PyTorch knowledge, distributed training experience, and strong English communication for collaboration across teams.

Qualifications

• 3–7 years of experience in applied research or R&D.
• Experience with large-scale training and distributed resources.
• Deep knowledge of PyTorch internals (memory, kernels, attention).
• Strong data pipelines, data quality, and versioning practices.
• Track record in research environments with high software standards.
• Fluent English; day-to-week collaboration in English.

Responsibilities

• Scale pre-training to new frontiers on transformers and long-context models.
• Fine-tune pretrained models on open-source and proprietary data.
• Engineer robust multimodal data pipelines for heterogeneous datasets.
• Enforce scientific rigor with internal leaderboards and data tests.
• Write production-ready research code: typed, tested, documented.

Skills

R&D experience

Distributed training

PyTorch mastery

Data engineering

Scientific rigor

English fluency

Job description

Sigma Nova is building foundational models for scientific signals (EEG, fMRI, ultrasound) and aims to extend to other domains. The architect of our Data & Training Factory will scale pre-training, fine-tune models, and engineer robust multimodal pipelines with a focus on data integrity and reproducibility.

The role requires 3–7 years in applied research, deep PyTorch knowledge, distributed training experience, and strong English communication for collaboration across teams.