The Story
Follow the narrative to build a mental model of how everything connects.
The Nerve Centre
At the heart of the AI campus sits the Azure AI Services hub -a single multi-service resource that provides one endpoint and one key to unlock vision, language, speech, and decision capabilities. Engineers authenticate with API keys or Azure AD tokens, monitor usage through Azure Monitor, and deploy containers to the edge when data cannot leave the premises.
The Visual Workshop
In the Visual Workshop, cameras analyse every frame. Image Analysis 4.0 (powered by Florence) generates captions, detects objects, and identifies people. The Read API extracts text from printed pages and handwritten notes alike. Across the corridor, Custom Vision trains specialised models to classify defects on a production line or detect safety gear on workers.
The Language Library
The Language Library buzzes with conversations. CLU (Conversational Language Understanding) replaces the old LUIS system, training on utterances to recognise intents and extract entities. Nearby, analysts run sentiment analysis and PII detection on customer feedback, while a Question Answering knowledge base fields common enquiries from URLs and documents.
The Voice Chamber
Voices echo through the Speech Chamber. Real-time speech-to-text transcribes meetings while custom acoustic models handle noisy factory floors. Text-to-speech with SSML controls pitch, rate, and emphasis for a natural narrator. Meanwhile, speech translation bridges languages in real-time, and speaker recognition confirms who is speaking.
The Document Forge
Stacks of invoices, receipts, and identity documents arrive at the Document Forge. Prebuilt models instantly extract fields from standard forms. For bespoke layouts, custom template models handle fixed forms while neural models adapt to varying structures. A composed model routes incoming documents to the right extractor automatically.
The Knowledge Mine
Deep in the Knowledge Mine, Azure AI Search indexes millions of documents. An indexer pipeline pulls from blob storage, runs built-in skillsets (OCR, entity recognition, key phrases), and enriches every document. Custom skills call Azure Functions for specialised processing. The enriched data lands in a searchable index -and optionally a knowledge store for downstream analytics.
The Generative Workshop
The Generative Workshop hums with creativity. Azure OpenAI models -GPT-4o, GPT-3.5-Turbo -respond to carefully crafted system messages that define persona and boundaries. The On Your Data feature grounds responses in enterprise knowledge via Azure AI Search. Content filters screen every prompt and completion for harmful content across four categories.
The Conversation Hall
Bots greet visitors in the Conversation Hall. Built with the Bot Framework SDK, they follow waterfall dialogs -step-by-step guided conversations with text, number, and choice prompts. The Azure Bot Service connects them to Teams, Slack, and web chat. Behind the scenes, CLU interprets intent while Question Answering handles FAQs.
The Ethics Council
Before any AI solution goes live, it passes through the Ethics Council. Six Responsible AI principles -fairness, reliability, privacy, inclusiveness, transparency, and accountability -guide every deployment. Transparency notes explain limitations. The Content Safety service screens for harmful content. Sensitive APIs require Limited Access approval, and high-risk decisions demand human oversight.
The RAG Assembly
In the final chamber, retrieval and generation unite. Documents are chunked, embedded, and indexed in Azure AI Search. When a user asks a question, hybrid search (keyword + vector) retrieves the most relevant passages. These are injected into the prompt for Azure OpenAI, which generates a grounded, contextual answer. The system message keeps the response on-topic and safe.
