Full-Stack Software Developer

Full-time Full-Stack Software Developer Sydney, Australia (Bankstown), Hybrid AUD$120k–$160k Apply online: https://www.ausrehab.com/careers/full-stack-ai-developer/ AusRehab is a workplace rehabilitation provider in Australia. We help injured workers in the NSW workers' compensation scheme return to work. We're replacing a stack of off-the-shelf systems with in-house, AI-driven ones, so our consultants can do their work better: fewer errors, statutory deadlines met, clearer communication with workers and insurers, and the audit and regulatory requirements handled properly rather than remembered at the last minute. For this we're assembling a small team of 3-4 full-stack developers who will work directly with our rehab consultants to design systems that fit the task. You'd own solutions end to end, with real autonomy and no ops team to hand your infrastructure to. You'd also talk to the people using what you build. No workers' comp knowledge needed, but we do ask for a curious mind to learn from an unfamiliar domain. [What the role covers] Agentic engineering — designing and implementing practical agentic solutions to real-world problems in the health industry. Not a pure research role, but some research and experimentation is expectedFull-stack web and mobile application development — front-end, back-end, database, telemetry, analytics, and reporting for in-house applications managing leads, cases, invoicing, communications, and auditsInfrastructure & DevOps — primarily AWS and GitHub (source control and CI/CD), with possible Docker/Kubernetes work ahead. Everyone owns their own infrastructure — no separate ops team to hand this off toDirect stakeholder engagement — working with Rehabilitation Consultants, Employment Consultants, Admin, and Accounts to understand, resolve, and train them on what gets built, alongside product management. Thinking on your feet and communicating directly with non-technical staff is part of the job The work is primarily integration and product engineering, connecting systems and processes that don't currently talk to each other. The person who thrives here can hold the full picture, move between layers without needing a specialist for every transition, and judge where to go deep versus move fast. [Tech stack] Python — for backend development and agentic workflows, at a production levelTypeScript and its ecosystem — React, React Native, Next.js for web and mobile; able to build and maintain meaningful features, not just CSS adjustmentsDatabase — relational database experience (PostgreSQL preferred; SQL Server/MySQL/Oracle also fine — the skill is highly transferable). Data lake experience is a bonusGit — good source control discipline is essentialInfrastructure — AWS and Kubernetes experience highly valued (Kubernetes implies Docker). The AWS services actually in use are IAM, S3, SQS, VPC, Lambda, and EC2 — not the full breadth of the platform. Good to have, but negotiableDevOps — GitHub Actions or equivalent CI/CD experience is a plus [Engineering skills] Strong communication — across cohorts: executives, fellow developers, and non-technical rehab specialists. Adjusting register and detail to the audience without dumbing down the substanceProblem-solving — reasoning from first principles rather than pattern-matching, running genuine trade-off studies, and knowing when to be rigorous (data provenance, ACL, billing logic) versus when to move fast — without being told. Comfortable working through ambiguity; requirements here are emergent, not fully defined ticketsTeamwork — fosters a collaborative spirit in-office and remote; reconciles different personalities and working styles; puts the team ahead of individual preference; owns responsibilities without being chased; strives to meet deadlines so the team moves in lockstep, flagging early when one is at risk [AI fluency] Understanding of key AI technologies — LLMs, context windows, tokens, prompt engineering, agents, MCP, skills/tools. Working literacy, not academic depthDesigning agentic processes to drive quality — engineering around LLMs' non-determinism and occasional hallucination while still harnessing their full power.Experience with an agentic framework — experience with LangChain, LangGraph, or comparable, is a big plus, though not essentialML experience is valued but not essential — much of the design thinking behind reliable agentic systems parallels classical ML practice (handling uncertainty, validating output, anticipating failure modes). Understanding transformer architecture helps with model selection, context window management, and anticipating LLM failure modesEffective use of an AI coding agent for coding, tooling, and experimentation is essentialCritical review of AI-generated work is essential — the judgment to assess designs, code, tests, and documentation, and say specifically why something is wrong or unsound, not just that it "feels off"