🛠 Tech Stack

💼 About This Role

You'll design and train large language models from scratch on Australian infrastructure, working directly with engineers on core architecture and training runs. Your work will shape Matilda, Australia's first LLM built from first principles. You'll spend substantial time in code, logs, and evaluation outputs, driving clarity on what improves the model.

🎯 What You'll Do

• Design and test architecture changes for large language models
• Run controlled experiments at scale and isolate causal effects
• Analyse training dynamics using logs, metrics, and model outputs
• Collaborate with ML systems engineers on distributed training

📋 Requirements

• Hands-on experience writing and running production-grade ML or research code
• Strong Python and experience with PyTorch or JAX
• Solid understanding of transformer-based language models and pre-training
• Ability to design experiments and interpret results

✨ Nice to Have

• Experience with distributed training concepts and tooling
• Familiarity with large-model training stacks like Megatron or DeepSpeed
• Experience working in ROCm-based environments

🎁 Benefits & Perks

• 🧠 Work on core LLM architecture from scratch
• 🔬 Access to one of Australia's largest private AI compute clusters
• 🤝 Close collaboration with engineers across the full stack