“If you are excited by the possibility that the next breakthrough in machine learning may come from outside machine learning itself, this is an opportunity to put that idea to the test. By bringing ideas from statistical physics to neural network compression, you'll explore whether we can move beyond trial and error toward discovering fundamental principles that govern how AI models scale. SPAARC offers you the chance to tackle ambitious scientific questions, challenge conventional thinking, and help lay the foundations for a new generation of efficient AI systems, with applications reaching all the way to space.”

- David Staszak, Lead Machine Learning Scientist, Advanced Technology

About the Role

Amii is seeking a Research Associate to contribute to SPAARC (Space-ready Principled Adaptive-AI via Renormalization-based Compression), a research project in collaboration with the European Space Agency (ESA).

The position focuses on developing physics-inspired approaches to neural network compression, with the goal of understanding how model architectures and performance evolve under increasingly constrained computational resources. The successful candidate will work closely with Amii research scientists as well as ESA scientists and take a leading role in the project's experimental research.

About the Project

SPAARC (Space-ready Principled Adaptive-AI via Renormalization-based Compression) investigates whether concepts from renormalization-group theory can provide a systematic and predictive framework for neural network compression. The project combines knowledge distillation, architectural optimization, and statistical analysis to study scaling relationships and performance limits in neural networks,



with applications to computer vision and space-based Earth observation.

Our primary partner on this project is the European Space Agency (ESA), an intergovernmental organization dedicated to advancing space exploration, scientific research, and innovative space technologies. Through its research and technology development programs, ESA supports emerging technologies that address the unique challenges of operating in space.

We are also collaborating with Mission Control, a Canadian space technology company specializing in AI and autonomous systems for space applications. Mission Control will contribute expertise in space-based AI and support the evaluation of compressed models for potential deployment on resource-constrained spacecraft.

Required Skills / Expertise

We are looking for a curious and motivated researcher with a robust foundation in physics and extensive hands‑on experience in machine learning. The ideal candidate is excited about exploring fundamental scientific questions, translating theoretical ideas into rigorous computational experiments, and developing innovative approaches to challenging research problems at the intersection of physics and AI.

Key Responsibilities:

- Contribute to developing the theoretical foundations for applying renormalization‑group (RG) concepts to knowledge distillation and neural network compression,



including formulating mathematical frameworks and testable hypotheses.
- Develop and evaluate machine learning models and algorithms for neural network compression and knowledge distillation.
- Design and conduct numerical experiments to investigate model scaling, architectural transformations, and compression limits.
- Develop research software and experimental pipelines using modern machine learning frameworks.
- Analyze experimental results, including statistical uncertainty, robustness, and sensitivity to architectural changes.
- Collaborate with Amii researchers and contribute to scientific publications, technical reports, and open‑source research tools.

Required Qualifications:

- PhD in Physics or Mathematics, with a strong foundation in statistical physics and renormalization group theory. Exceptional candidates with different backgrounds (e.g., a Master’s) with demonstrated expertise in model compression or knowledge distillation will also be considered.
- Extensive experience in machine learning, including designing, training, and evaluating deep neural networks.
- Strong Python programming skills and proficiency with modern machine learning frameworks, particularly PyTorch.
- Familiarity with computer vision concepts and techniques, including image classification, object detection, and segmentation. Hands‑on experience in these areas is considered an asset.
- Experience with statistical analysis, uncertainty quantification, and reproducible computational experiments.
- Familiarity with GPU computing, Linux, and Git.
- Demonstrated ability to conduct independent research, develop computational methods, and perform rigorous quantitative analysis.
- Strong scientific communication skills and

📌 Research Associate (Edmonton)
🏢 Alberta-Machine-Intelligence-Institute-1
📍 Edmonton