Services

LLM Applications & RAG in Adelaide

Production-grade large language model applications with retrieval-augmented generation, evaluation and security.

Discuss your projectBook a consultation or learning sessions

LLM Applications & RAG for South Australian organisations

Large language models can read, summarise, classify and generate text at a scale no team can match. Moving from an impressive demo to a reliable production system is a different challenge, though. It takes careful data engineering, evaluation, security, cost management and attention to data sovereignty.

EITSAA engineers LLM applications for South Australian organisations that need them to work every day. We design retrieval-augmented generation (RAG) systems over your private data, build document processing pipelines, and put rigorous evaluation in place, so quality is measured rather than assumed.

A demo is easy. A production LLM application needs evaluation, cost control, security and data sovereignty. We engineer all four from day one.

What's included

  • Retrieval-augmented generation (RAG) over private data
  • Document processing: summarisation, extraction, classification
  • Model selection across AWS Bedrock, Azure OpenAI and open-weight models
  • Evaluation, red-teaming and prompt-injection defences
  • Data kept in Australian cloud regions where required
  • Online libraries for organisations and schools, with LLM search, GraphQL queries and connectors to SSO, SharePoint, learning management and records systems

Illustrative demo

Watch a document become an answer

See how a retrieval-augmented generation (RAG) pipeline works, from a new file arriving to a cited answer. This is a scripted demonstration with a fictional council policy and sample data. It makes no network calls.

MLOps

MLOps on AWS: from data to a monitored model

We use managed AWS services so models are built, tested, approved and watched the same way every time. We select only the services your workload needs. Each stage produces evidence for AI governance.

  1. Data and feature prep

    Versioned, access-controlled data and features, kept in an Australian region.

    • Amazon S3
    • AWS Glue
    • SageMaker Feature Store

  2. Training and fine-tuning

    Repeatable training or fine-tuning runs, only where retrieval and prompting are not enough.

    • SageMaker AI training
    • Amazon Bedrock customisation

  3. Evaluation

    Quality, safety and bias tests against a fixed test set before anything is promoted.

    • Amazon Bedrock Evaluations
    • SageMaker Clarify

  4. Registry and approval

    Every model version recorded with its lineage, results and a named human approver.

    • SageMaker Pipelines
    • SageMaker Model Registry

  5. Deploy

    Automated, reviewed releases with rollback, for hosted models and RAG knowledge bases.

    • AWS CodePipeline
    • Step Functions
    • SageMaker AI endpoints
    • Bedrock Knowledge Bases

  6. Monitor and drift

    Track quality, cost, latency and drift, with alerts routed to an owner.

    • SageMaker Model Monitor
    • Amazon CloudWatch

  7. Retrain loop

    Drift or new data triggers the pipeline again, and the new version goes back through evaluation and approval.

    • Amazon EventBridge
    • Step Functions

Stage 7 returns to stage 1: the loop repeats whenever monitoring or new data calls for it.

Online library

An online library for your organisation or school

We build a searchable library of policies, resources and records with an LLM-powered search bar. Questions become GraphQL queries over your approved sources, and answers cite what they came from. It connects to the systems schools, councils and agencies already use.

  1. SourcesSharePoint and M365, document stores, learning management and records systems
  2. ConnectorsScheduled and event-driven sync with permissions kept
  3. IndexText and vector index with metadata
  4. GraphQL APIOne typed API for search, browse and records
  5. LLM answerGrounded, cited and filtered by who is asking
  6. Library UISearch bar, filters and accessible results
  • Identity and SSO (Entra ID, Google Workspace)
  • Student or records systems
  • SharePoint and Microsoft 365
  • Learning management systems
  • Document stores and file shares
  • Library catalogues

Discuss this

How we can help

Retrieval-augmented generation

Question answering over large document collections, with citations, permission-aware retrieval and freshness controls.

Document intelligence

Extract structured data from forms, invoices, contracts and reports, with human review queues.

Summarisation and drafting

Generate briefs, reports and correspondence drafts in your organisation's style and format.

Classification at scale

Route, tag and prioritise thousands of records cost-effectively.

Evaluation and red-teaming

Test suites that measure accuracy and safety and defend against prompt injection before and after release.

Our delivery process

  1. Discover

    We clarify goals, users, constraints and success measures, and review your current systems and data.

  2. Design

    We agree the architecture, security controls and user experience before the build begins.

  3. Deliver

    Fortnightly sprints with demos, automated testing and continuous deployment.

  4. Operate and improve

    Monitoring, support and a backlog of improvements based on real usage.

Technology we use

AWS Bedrock, Azure OpenAI, open-weight models, OpenSearch, pgvector, Python, LangChain, LlamaIndex, Ragas and custom evaluation harnesses.

Why choose an Adelaide partner?

Working with a local team means face-to-face workshops, an understanding of South Australian regulatory and procurement requirements, and support in your time zone. Engaging EITSAA also keeps investment and jobs in South Australia, which counts under the SA Industry Participation Policy in government tenders.

See our government capability statement or read how we work.

Outcomes you can expect

Every LLM Applications & RAG engagement starts with outcomes we agree on, not just a list of features. Before we begin, we set measures such as hours saved, faster turnaround, higher customer satisfaction, lower operating cost or reduced risk. We report against them throughout delivery. You get a fortnightly demonstration of working software, a transparent backlog, and a named Adelaide-based lead who is accountable for progress.

When the work is done, you keep the capability. Documentation, source code, infrastructure definitions and runbooks are handed over in full, and knowledge-transfer sessions give your team the confidence to run and extend what we have built.

Security, privacy and compliance

Security and privacy are designed in from the first workshop. Our LLM Applications & RAG solutions follow the principle of least privilege, encrypt data in transit and at rest, and can be hosted in Australian cloud regions to meet data sovereignty requirements. We align with the ACSC Essential Eight, the Australian Privacy Principles and, for public sector clients, the SA Government Information Security Management Framework.

For regulated or sensitive work, we can complete privacy impact assessments, provide security documentation for your risk team, and support independent penetration testing before go-live.

Frequently asked questions

Should we use RAG or fine-tuning?

For most business knowledge use cases, RAG is the better starting point. It keeps answers current and citable. Fine-tuning suits narrow, consistent tasks. Read our RAG vs fine-tuning guide.

Can our data stay in Australia?

Yes. We can deploy on AWS Sydney or Melbourne regions and Azure Australia regions, and we use models that support local data residency.

How do you measure quality?

We build an evaluation set of real questions and expected answers, then track accuracy, faithfulness and safety on every change.

How do you control costs?

Through model selection, caching, prompt optimisation and usage monitoring, with cost dashboards from day one.

How do you run MLOps on AWS?

We use managed AWS services to prepare data, train or fine-tune where it is justified, evaluate, approve models in a registry, deploy, and monitor for drift and quality. Services typically include Amazon S3, Amazon SageMaker AI, Amazon Bedrock, Step Functions, EventBridge, CodePipeline and CloudWatch, chosen to suit your workload.

Can you build an online library with an AI search bar?

Yes. We build a searchable library over your approved content. The search bar turns a question into a GraphQL query, retrieves matching sources the person is allowed to see, and returns an answer with citations. It connects to identity and SSO, SharePoint and Microsoft 365, learning management systems, records systems and document stores.

Further reading

Plain-English guides from our Insights library related to LLM Applications & RAG.

Browse all insights →

Learn

Reference architectures for LLM Applications & RAG

Related services

AI Governance & Assurance

Governance, guardrails and assurance for every kind of AI: machine learning, generative AI, LLMs, agents and automated decisioning, aligned to Australian guidance and international standards.

Explore AI governance →

Training & Team Consultations

Workshops, teaching sessions and expert consultations that lift your team's skills in AI, software engineering, cloud, data and security, tailored to your own tools and use cases.

Explore training →

Talk to us about LLM Applications & RAG

Tell us what you need. We reply within one business day and the initial conversation is free of charge.

Discuss your projectBook a consultation or learning sessionsCall 0402 401 584