Forward-Deployed AI Engineer, Justice and Community Safety

Maincode makes Matilda, an Australian-made AI stack we own end to end, from tin to token. We own and operate our own infrastructure here in Australia, we train and maintain our own frontier models, and we run multiple surfaces on top of them, including chat, coding and API. Everything we ship is built on capability we control ourselves.

We are hiring someone who cares about that. You should be passionate about Australian-made AI, a builder first, and keen to work directly with customers to help them move forward on the Matilda stack.

You will be Maincode's senior technical presence with customers in Melbourne, including departments across the Justice and Community Safety (JACS) portfolio.

This is a forward-deployed engineering role. You will embed with customer teams, get to grips with their hardest data problems, and design, adapt and deploy the Matilda stack to solve them. You will be the technical face of Maincode in the Justice and Emergency Services (J&ES) community: credible in a room of senior technical staff, able to prototype in days, and rigorous enough to ship working systems into classified environments.

• →

  Embed with Victorian J&ES customers to find where Matilda can deliver real mission outcomes, and turn loosely defined requirements into concrete technical plans.

• →

  Design, fine-tune, evaluate and deploy systems on the Matilda stack against customer data, including in on-premises, air-gapped and classified environments.

• →

  Own delivery end to end, from data pipelines and fine-tuning through retrieval-augmented generation, evaluation, MLOps and hardening to relevant Government security standards.

• →

  Brief confidently at every level, from working engineers through to executives, and write architecture papers and evaluation reports that stand up to scrutiny.

• →

  Bring what you learn in the field back to the team that trains the models, so customer experience directly shapes the next version of the stack.

Essential

• →

  Must hold an active AGSVA NV1 (or higher) security clearance and be willing to undergo Victoria Police organizational suitability vetting.

• →

  Deep hands-on experience in data science and ML engineering, with strong Python, modern ML stacks such as PyTorch, JAX and the Hugging Face ecosystem, and solid software engineering fundamentals.

• →

  Demonstrated experience deploying ML systems in secure government, defence or intelligence environments, including on-premises and air-gapped infrastructure.

• →

  Working knowledge of the Victorian Protective Data Security Framework (VPDSF), the ISM, the Protective Security Policy Framework (PSPF), and data classification and handling requirements.

• →

  Excellent judgement and discretion, and the ability to operate autonomously inside customer environments.

• →

  The ability to explain complex ML concepts crisply, and to write exceptionally well.

Highly Regarded

• →

  Prior experience within State Law Enforcement, Public Safety, or Federal Intelligence (e.g., VicPol, ASIO, ASD, or Defence S&T), ideally in a data science, technical, or analytic role.

• →

  Hands-on experience with large language models: fine-tuning, evaluation, retrieval-augmented generation, and their known failure modes.

• →

  Familiarity with intelligence-adjacent data workflows such as entity resolution, geospatial analysis, OSINT, graph analytics, or tradecraft tooling.

• →

  Solutions architecture or technical pre-sales experience in a commercial setting.

This position requires an active AGSVA NV1 (or higher) clearance. Due to clearance eligibility requirements, applicants must be Australian citizens. You will be required to consent to AGSVA revalidation, and to obtain and maintain customer-specific briefings and accesses as required.

Builders who want to spend time with customers, and customer people who actually build. If you care about Australian-made AI and you would rather spend your week deploying systems with agencies than talking about them, you will do well here.