The AI team is comprised of highly experienced designers, developers, data scientists and analysts who are responsible for delivering digital products and AI solutions to PwC’s business across multiple lines of service in Australia.

We partner with subject matter experts across the business to advise on, design and build agentic AI systems and LLM-powered products that transform the way we deliver assurance, advisory, tax and legal services to our clients.

As an AI Engineer / Data Scientist on our team, you will design, build and evaluate production agentic AI systems, LLM applications that reason, call tools, retrieve knowledge and complete multi-step tasks on behalf of our practitioners and clients. You will own problems end to end: from framing the approach and prototyping, to shipping reliable, observable agents at scale in the Azure cloud.

We are a data science team at heart, so we care deeply about rigour: measuring agent quality, running evaluations, and using the scientific method to know whether something actually works, not just whether it demos well.

What you’ll do:

• Building agentic solutions: agents that plan and execute multi-step tasks over long horizons, with context compaction and tool-call repair so they stay reliable and on-budget.

• Designing and expanding agent tool ecosystems: file/document operations, sandboxed code and bash execution, web search, and enterprise integrations (e.g. MS Graph) — with well-typed, testable schemas.

• Orchestrating multi-agent / sub-agent systems that spawn parallel sub-tasks, coordinate, and report results back to a lead agent.

• Building stateful LLM workflows (e.g. LangGraph) for domain pipelines such as research, summarisation and report generation.

• Engineering RAG and retrieval pipelines: chunking, embeddings, vector search (pgvector / Azure AI Search), hybrid and re-ranked retrieval, and grounding LLM outputs in trusted sources.

• Producing structured, reliable LLM outputs (function calling / Pydantic) and designing robust prompts and agent skills.

• Building evaluation harnesses and offline batch evals (including deep-research style runs) to measure accuracy, faithfulness, cost and latency — and using those signals to iterate.

• Instrumenting agents with observability and tracing (e.g. Langfuse, OpenTelemetry) and tracking token usage and cost in production.

• Routing across multiple models and providers via an enterprise gateway (e.g. LiteLLM) and tuning model selection for quality, latency and cost.

• Producing clean, maintainable, efficient code deployed at scale in the Azure cloud; scaffolding new projects, pairing with engineers, and reviewing pull requests.

• Contributing to team stand-ups and the broader software development lifecycle, and participating in firmwide data science, ML and AI forums.

• Coaching and mentoring junior data scientists and engineers.

What we’re looking for: Applicants must be able to demonstrate the following key capabilities. We do not expect every candidate to tick every box, strong fundamentals and a track record of shipping LLM-powered products matter most.

Core skills:

• Strong Python development experience, with hands-on use of modern LLM frameworks and SDKs (e.g. OpenAI / Anthropic SDKs, LiteLLM, LangGraph).

• Deep, practical knowledge of prompt engineering, LLM workflows and agentic patterns, tool/function calling, multi-step agents, and structured outputs.

• Experience building and optimising RAG pipelines, including evaluation, with industry-standard tooling.

• Working knowledge of vector databases such as pgVector and Azure AI Search, plus embeddings and hybrid/semantic search.

• Strong experience with SQL databases such as PostgreSQL (or equivalents).

• A scientific, evaluation-first mindset: selecting appropriate methods, applying algorithms at scale, and using the scientific method to derive robust, defensible conclusions about model and agent behaviour.

• Strong critical thinking, analytical rigour and outstanding attention to detail.

• Proper source code management and confident use of Git.

• Good written and verbal communication, and the ability to work effectively with remote teams.

• A proactive, problem-solving approach and the ability to solve complex problems as part of a team.

Desirable skills:

• Experience with LLM observability and cost/quality tracing (e.g. Langfuse, OpenTelemetry).

• Experience routing across multiple models/providers (e.g. via LiteLLM or an enterprise gateway) and reasoning about model selection trade-offs.

• Experience designing multi-agent / sub-agent systems and agent tool ecosystems (web search, file/document tools, enterprise integrations such as MS Graph).

• Experience with computer-use / browser-automation agents (e.g. Playwright).

• Knowledge of classical ML (regression/boosting) and deep learning (CNN/RNN), preferably in NLP or CV.

• A research background in ML/LLM model development, and the ability to identify emerging techniques and apply them to practical situations.

• Experience with microservices, containerisation (Docker), and building/operating data pipelines at scale.

• Experience with message-queueing solutions (e.g. RabbitMQ, Kafka).

• Experience developing on cloud environments, particularly Azure (Azure OpenAI, AI Search, Blob, Key Vault, App Insights).

• Knowledge of agile software development lifecycles (SDLC) and experience on agile projects.

• TypeScript / full-stack experience is a strong bonus: much of our agent tooling and UI is built in TypeScript (Vercel AI SDK, NestJS, Next.js / React), and being able to work across the Python and TypeScript stack is highly valued.