Maincode is building an AI model factory. We’ve built Matilda, Australia’s first large language model trained entirely from scratch, and we’re developing new reasoning architectures beyond LLMs. We design and train models from first principles, deploy them at scale, and push the limits of what can be built here in Australia.

As an AI Engineer, you’ll help make this possible. You’ll work on both sides of the stack: how models are trained, and how they run in the world. You’ll design and build systems for large-scale training and efficient inference, collaborating closely with AI Researchers to turn new ideas into production-grade systems.

This is a deep engineering role. You’ll write a lot of code, instrument distributed systems, optimise performance, and debug the complex, messy edge cases that come with large-scale model training and deployment.

• Build and scale training pipelines for large language models and other architectures

• Design and optimise inference systems for real-world deployment

• Improve data pipelines, tokenisation, and distributed training across multi-node clusters

• Build observability into training and inference to improve speed, efficiency, and correctness

• Diagnose and resolve distributed training and GPU-related issues

• Work closely with AI Researchers to translate cutting-edge algorithms into robust systems

• Establish ModelOps practices for reliability, reproducibility, and continuous improvement

• Deeply interested in how models are built, trained, and run

• Curious about model internals, training dynamics, and system performance

• Comfortable working across both training and inference stacks

• Skilled in Python and frameworks like PyTorch or JAX (familiarity with CUDA, Triton, or NCCL is a plus)

• Hands-on, persistent, and motivated by solving hard problems

• Constantly learning: reading papers, exploring repos, or experimenting with new tools

• Excited to help shape the next generation of Australian AI capability

We’re a small, focused team building advanced AI systems from the ground up. We run our own GPU clusters, train our own models, and collaborate closely across research and engineering to move fast and build real capability.

• Care about how models actually work, not just what they output

• Build things that run at scale

• Value curiosity, rigour, and craftsmanship

• Want to help Australia develop world-class AI systems