You will build the 'brain' of tactical robots by designing large-scale, multi-modal foundational models that learn from unlabelled Electro-Optical and Infrared data
Job Summary
You will build the 'brain' of tactical robots by designing large-scale, multi-modal foundational models that learn from unlabelled Electro-Optical and Infrared data.
The role requires managing and optimizing distributed training pipelines across multi-node GPU clusters while developing metrics to validate latent space representations.
Harmattan AI is a next-generation defense prime seeking candidates with a hybrid researcher-engineer mindset to deliver mission-critical systems to allied forces.
Matching Summary
You will build the 'brain' of tactical robots by designing large-scale, multi-modal foundational models that learn from unlabelled Electro-Optical and Infrared data.
Skills & Requirements
Must-have
Self-Supervised Learning architecture design
Multi-modal EO and IR data processing
Vision Transformers and Masked Autoencoders
Distributed training on multi-node GPU clusters
PyTorch engineering and mathematical intuition
Nice-to-have
Experience with C++, Rust, or Go languages
Edge AI model distillation knowledge
Cross-attention mechanism implementation
Fault-tolerant data pipeline architecture
Ethical defense technology mindset
Key Requirements
PhD or research-focused MS in Computer Science or related field
5-6 years of experience training deep learning vision models
Proven application of SSL architectures to non-standard imaging data