Sydney, Hybrid, 3 days per week in a Sydney CBD office
$130,000 to $180,000 plus super, junior to senior levels
Our client is a large global professional services organisation building out an in house AI team that designs, builds and evaluates production agent systems. This is not a chatbot wrapper team. Their agent harness is built in TypeScript on the Vercel AI SDK, with model and provider routing through a LiteLLM based gateway, running at scale in Azure. Python is used heavily alongside TypeScript for evaluation, data work and model experimentation, so comfort moving across both stacks is genuinely valued.
They are evaluation led. Every decision is backed by measurement, not demos. They strongly prefer candidates who have worked at startups and bring that vibe, people who are comfortable moving fast and owning problems end to end, even though the business behind the team is large.
What you'll do
Build and extend an agent harness in TypeScript including the runtime loop, tool dispatch, retries, tool call repair, and state persistence
Design context and memory strategies for long horizon agent work including compaction, caching and token budgeting
Architect multi agent and sub agent systems with lead and worker patterns, delegation and shared state




Engineer RAG pipelines on PostgreSQL with pgvector, including hybrid and re ranked retrieval
Build evaluation harnesses and trajectory level assessment so quality is proven rather than assumed, using Python for evaluation runs, data work and model experimentation
Instrument agents with observability tooling such as Langfuse and OpenTelemetry
Mentor junior engineers as part of a scaling team
What we're looking for
Proven experience shipping production AI systems. This is essential, not just prototypes, POCs or internal demos
Working Python capability, including evaluation, data work and model experimentation. Comfort moving across TypeScript and Python day to day is important for this role
A background at a startup or strong affinity with startup pace and culture
Strong TypeScript experience, confident with the type system for tool schemas and structured outputs
Hands on experience building agents with a contemporary framework, Vercel AI SDK strongly preferred
Practical context engineering experience including caching and cost and latency trade offs
Senior candidates should bring multi agent architecture experience and be comfortable mentoring others
#J-*****-Ljbffr

📌 Artificial Intelligence Engineer (New South Wales)
🏢 Novus Pathway
📍 New South Wales