AI Developer

Salary: $150,000 – $170,000 per year

AI Developer – Local AI Agents for Legal Workflow Automation

R&R · Kogarah, Sydney NSW · Full-time · On-site / WFH

About the role

We're a legal practice building our own private AI capability, and we want an experienced AI developer to implement it. You'll design, train and run specialised AI agents on our local AI workstation. They'll classify incoming work and drive our workflows. Nothing leaves our environment — client confidentiality comes first.

Your models will sit behind APIs used by our .NET desktop and web applications. Those apps take in tickets, emails, and phone calls, pass them to your agents, and write the results to our database. You'll own the AI layer end to end and work closely with our in-house development team.

What you'll do

• Build, fine-tune, and evaluate local language models and classifiers for legal work, such as matter type, urgency, document type and routing.

• Design agent workflows that extract key details from emails, tickets, and call transcripts and trigger the right next step.

• Prepare and label training data from our historical records, with proper handling of privileged and confidential information.

• Expose models and agents through secure, well-documented REST APIs for our .NET applications.

• Integrate with our SQL databases so agents can look up and update matter, client, and workflow records.

• Set up and maintain the AI stack on our Ubuntu workstation: GPU drivers, model serving, versioning, monitoring and backups.

• Measure accuracy, add human-in-the-loop review where confidence is low, and keep audit logs of AI decisions.

• Document your work so the wider team can support and extend it.

Our environment

• AI platform: a dedicated, well-stocked, on-premises AI workstation running Ubuntu, provided by us.

• Applications: .NET desktop and web apps (.NET Core / .NET and .NET Framework) that act as the middle layer between incoming requests and our database.

• Data: SQL Server databases holding case, client and workflow records.

• Inputs: tickets, emails and phone calls (transcribed).

• Principle: local-first AI. Client data stays on our infrastructure.

What you'll need

• At least 3 years of proven commercial experience building and deploying AI/ML solutions, with examples you can show or talk through.

• Strong Python, including PyTorch and Hugging Face Transformers (or similar).

• Hands-on experience running and fine-tuning open-weight LLMs locally (for example Llama, Mistral or Qwen) using tools such as Ollama, vLLM or llama.cpp.

• Experience with decision models such as Jev, Kev, other system one type models and APIs.

• Experience building agent or workflow automation that calls tools, APIs and databases.

• Familiarity with .NET Core / .NET and .NET Framework (C#), enough to design APIs our apps can use and to work inside our codebase.

• Solid SQL skills.

• Comfort administering Linux (Ubuntu), including GPU/CUDA setup, Docker and service management.

• Building secure REST APIs (for example FastAPI) with authentication and logging.

• A careful approach to confidential data, privacy and security.

• Exceptional written and verbal communication skills across all organizational tiers.

Nice to have

• Experience in legal, professional services or another regulated industry.

• Speech-to-text (for example Whisper) for call transcription.

• Microsoft 365 and Microsoft Graph integration (email, Teams, SharePoint).

• Fine-tuning methods such as LoRA/QLoRA, and model quantisation.

• MLOps tooling for experiment tracking and model versioning (for example MLflow).

• Knowledge of the Australian Privacy Principles and legal professional privilege.

What we offer

• Competitive salary + super

• The chance to build a private AI platform from the ground up, with real ownership.

• Dedicated local AI hardware to work with.

• WFH, training budget

How to apply

To apply, please submit your CV and a short cover letter. Tell us about one AI project you've taken from idea to production: what it did, how you built it and how you measured it. Links to GitHub, a portfolio or demos are welcome.

Candidates must have permanent work rights in Australia.