Vocabulary: Claiming the Words
This lesson adds no new knowledge. Every term below names something you already do or have already drilled — but if you never attach the label to the experience, a panel scanning its answer key for the label marks the box empty. Each entry maps a checklist word onto work you can already speak to, with the claiming sentence ready to say.
Natural Language Processing (NLP) — the AI domain of understanding and generating human language: extraction, summarization, question answering, embeddings, RAG. Lives in: everything you've drilled — RAG, embeddings, grounded generation are NLP. Say it: "The retrieval-augmented systems I build are applied NLP — embedding documents, retrieving by meaning, and generating grounded, cited answers from them."
Computer Vision — the AI domain of extracting meaning from images and documents: OCR, document layout analysis, object detection, image classification. Lives in: the Azure workloads family; for the Commission, the document-intake front door. Say it: "Our vision use case is document intelligence — OCR and layout analysis on scanned transcripts, turning an image into structured coursework data that the NLP pipeline can then reason over. Azure Document Intelligence is the managed service for exactly that."
Predictive Analytics — using historical data to predict outcomes: workload forecasting, processing-time estimates, anomaly detection. The umbrella business term for the classical ML you already know — regression predicts the number, classification predicts the category. Lives in: classical ML (module 13, lesson 5 — same machinery, business label). Say it: "I've piloted predictive analytics in production — statistical anomaly detection using median and MAD baselines on our batch-run history, flagging runs that succeed but behave abnormally. Scaling that from statistics to a trained model is a supervised-learning problem on labels we already collect."
SDLC (Software Development Lifecycle) — the phases software moves through: requirements, design, build, test, deploy, operate. The interview point is that AI features ride the same lifecycle — plus evaluation gates. Lives in: software engineering practice; the frame that makes AI work sound governed. Say it: "I run the full lifecycle solo on our monitoring platform — requirements from operations staff, design, build, test, deploy as a Windows service, then operate and monitor. AI features add one phase: an evaluation gate before release, and re-running the evals on every model or prompt change."
Automation — replacing manual, repeatable work with scripted, scheduled, monitored processes. Lives in: your strongest ground — claim it by name, not just by example. Say it: "Automation is the core of my current role: I own the Commission's thirty scheduled batch jobs end to end — an estate I modernized from legacy Perl to PowerShell 7 — and I automated the monitoring of the automation itself, with a thirty-three-check daily health scan."
The bridge sentence — the one the tip at the bottom of every prep guide asks for: combining traditional .NET development with modern AI integration. Lives in: your portfolio close; have it ready before they ask. Say it: "I've already combined the two in production: a .NET-style full-stack platform — typed models, REST APIs, server-side views — with an AI diagnosis feature integrated through an OpenAI-compatible API, grounded in our runbooks, gated by role-based access and a feature toggle. The AI is a component inside a governed application, not a bolt-on."
The claiming drill
Once through the list daily: read only the bold term, say the claiming sentence from memory, then check. The test isn't whether you know the concept — you do — it's whether the label arrives welded to your evidence within the first breath of the answer.