Machine Learning Engineer - Foundational

Harmattan AI seeks a Senior Machine Learning Engineer for its Foundational team in Zürich, focused on developing large-scale, multi-modal models from unlabelled EO and IR data to power tactical robots. The role requires deep expertise in self-supervised learning, PyTorch, and scalable deep learning systems.

This role is part of the Foundational team, responsible for building the core intelligence for tactical robots. The engineer will design and scale multi-modal foundational models using self-supervised learning from unlabelled electro-optical and infrared data, providing essential weights that the Edge AI team distills into high-accuracy models for deployment on tactical hardware.

Responsibilities

• Create neural network architectures such as Vision Transformers and design loss functions like Masked Autoencoders and Contrastive Learning to learn from paired and unpaired EO and IR data.
• Oversee and optimise training pipelines on multi-node GPU clusters, including mixed-precision training and efficient data loading.
• Develop evaluation metrics and linear-probing benchmarks to validate that the model’s latent space captures meaningful semantic features before distillation.
• Review existing EO/IR data lakes and implement cross-attention mechanisms to integrate features from different sensor types.
• Coordinate with Data Engineers on data ingestion pipelines and work with the Edge AI team to ensure smooth, high-performance model transfers.

Requirements

• A PhD or a research-intensive Master’s degree in Computer Science, Machine Learning, Computer Vision, or Applied Mathematics.
• At least 5–6 years of experience at a senior level.
• Proven experience training and scaling deep learning vision models (ViTs, CNNs) from scratch in multi-GPU or multi-node environments.
• Demonstrated success in applying novel self-supervised learning or multi-modal architectures (e.g., CLIP, MAE, DINO) to real-world, non-standard imaging data such as IR, SAR, or hyperspectral.
• Strong PyTorch engineering skills combined with deep mathematical understanding of representation learning.
• Familiarity with system-level languages (C++, Rust, or Go) and resource optimisation for edge computing.
• Ability to design state machines for fault-tolerant data pipelines and manage technical trade-offs between hardware and algorithm teams.
• Full commitment to Harmattan AI’s mission of delivering an ethical defence advantage to allied nations.
• A hybrid researcher-engineer mindset that prioritises data quality as much as algorithm design.

About the company

Harmattan AI is a next-generation defence technology company developing autonomous and scalable defence systems. Valued at $1.4 billion after a $200M Series B round, the company focuses on impactful innovation, excellence, ambitious goals, and tackling the most challenging technical problems in a rigorous, high-performance environment.

• Commitment to building technologies with real-world impact.
• Pursuit of excellence and setting ambitious technical goals.
• Focus on solving the hardest technical challenges.
• Expectation of rigor, ownership, and execution in a demanding environment.