AI courses: learn AI by doing
162 interactive lessons — each one a narrated presentation with predict-and-check exercises, hands-on builds and hand-picked videos. Start with the shared Essentials, then follow the track for your role.
Pick a track
Prompt Mastery Track
Write prompts that get better answers — measurably
Learn the handful of habits that separate a vague request from a prompt that returns work you can actually use: role, context, constraints, format and examples. Practise them in the Prompt Lab on real work scenarios — it takes a few hours…
Engineering Track
Build with AI — code, APIs, RAG, agents, evaluation
Go from calling an LLM API to shipping something you would put in front of customers: structured outputs, embeddings and RAG, agents and tool servers, evaluation, tracing, cost control and security. Choose this if you write, test or…
Business & Builder Track
Use AI daily — prompts, workflows, no-code automation
Turn AI into an everyday work habit — drafting, summarizing, analysing spreadsheets and research, plus role playbooks for PM, QA, BA, HR, marketing and support, no-code automation and the privacy rules that keep you safe. Choose this if…
AWS AI Track
Build AI on AWS — Bedrock, SageMaker, agents and the AI Practitioner cert
Learn the AWS generative AI stack end to end: foundation models through Amazon Bedrock, custom models on SageMaker AI, RAG with Knowledge Bases, production agents on AgentCore, and the guardrails and governance a security review will ask…
Azure AI Track
Build AI on Azure — Microsoft Foundry, AI Search, agents and AI-901/AI-103
Work through Microsoft's AI platform as it stands today: Microsoft Foundry (formerly Azure AI Foundry), the Foundry model catalogue, grounding with Azure AI Search, prompt and hosted agents, Content Safety, and Entra-based governance…
Google Cloud AI Track
Build AI on Google Cloud — Gemini, Vector Search, ADK agents and BigQuery ML
Cover Google Cloud's AI surface as reorganised in 2026: Gemini and Model Garden, Vertex AI Vector Search, grounding and citations, the open-source Agent Development Kit, BigQuery ML and responsible AI controls. Choose this if your data…
NVIDIA AI Track
Run AI on your own hardware — CUDA, NIM, NeMo, guardrails and NCA certs
Go one layer below the managed clouds: how GPUs and VRAM actually constrain AI, deploying models as NIM microservices, training and fine-tuning with NeMo, self-hosted RAG with NeMo Retriever, and programmable safety with NeMo Guardrails…
MCP Builder Track
Learn MCP, build a server, ship it — then publish it for everyone else
The Model Context Protocol is how an assistant reaches a tool without an integration per pair. This track teaches the 2026-07-28 revision specifically, because it removed the initialize handshake, sessions and the SSE stream that almost…
How the levels work
- Essentials — The shared floor. What a model is, tokens and context, prompting that works, checking output, and the privacy rules. Everyone starts here.
- Foundation — The Essentials plus your track's orientation — enough to be useful and to know what you are looking at.
- Practitioner — The working middle of your track: the things you will actually build, configure and run.
- Advanced — Production concerns — scale, cost, evaluation, security, governance and the capstone.
- Go deeper — Optional library. Researched in full, browsable any time, and deliberately not counted toward your track or required to finish it.
AI Foundations & Concepts
The core ideas behind modern AI — machine learning, neural networks, transformers, and the ethical ground rules. Start here to build intuition everything else stands on.
- Introduction to Generative AIDiscover what generative AI is, how it differs from traditional software, and why models like ChatGPT can write, code…30 min · beginner
- Machine Learning FundamentalsLearn the core ideas behind all of AI: how machines learn from data through training, loss functions, and…45 min · beginner
- Neural Networks BasicsUnderstand neurons, layers, weights, and backpropagation — the building blocks that power everything from image…45 min · beginner
- Transformers ArchitectureMeet the architecture behind ChatGPT, Claude, and Gemini — and the attention mechanism that changed AI forever.1 hours · intermediate
- Natural Language ProcessingHow computers turn messy human language into numbers they can reason about — from tokens and embeddings to modern LLM…40 min · beginner
- Computer Vision FundamentalsHow machines see: from pixels to convolutional networks to vision transformers that recognize faces, tumors, and stop…40 min · beginner
- Reinforcement Learning OverviewLearning by trial and reward — the paradigm behind game-playing AIs and the RLHF technique that made ChatGPT helpful.40 min · intermediate
- Ethical AI PrinciplesThe core principles — fairness, transparency, privacy, and accountability — that separate beneficial AI from harmful AI.30 min · beginner
- Embeddings & Vector Databases in PracticeUnderstand embedding spaces, similarity metrics, ANN indexes, and how to choose and operate a vector store.50 min · intermediate
- Evaluating LLM Applications: Golden Sets, Judges & Regression TestsBuild the measurement layer: task metrics, curated golden sets, LLM-as-judge with calibration, and RAG-specific…1 hours · advanced
- Spotting Hallucinations and Fact-Checking AI OutputAI can state wrong things with perfect confidence. Learn where errors cluster, the red flags that signal one, and a…35 min · beginner
- AI for Spreadsheet and Data Analysis Without FormulasAsk questions of your data in plain English, get formulas and pivot logic explained, and learn the checks that stop a…40 min · intermediate
- AI for Research and Competitive AnalysisRun faster market, vendor and competitor research using AI as a structuring and synthesis partner — with source…40 min · intermediate
- The AWS AI Certification Path: AIF-C01 and What Comes AfterMap the AWS AI credentials — AI Practitioner, ML Engineer Associate, AI Business Strategist — and pick the one that…40 min · beginner
- AWS Managed AI Services: Textract, Comprehend, Rekognition and FriendsThe pre-trained API layer that solves document, text, image and speech problems in an afternoon — and the cases where…45 min · beginner
- The Microsoft AI Certification Ladder: AI-901 and AI-103Microsoft renumbered and renamed both Azure AI certifications. Learn which exam exists today, what it actually…40 min · beginner
- Foundry Tools: Vision, Speech, Language and TranslatorThe pre-trained, non-generative Azure AI services - now branded Foundry Tools - that solve a large share of real…45 min · beginner
- Azure AI Document Intelligence and Content UnderstandingTurn scanned forms, invoices and contracts into structured data with confidence scores - and know when to reach for…50 min · intermediate
- Google Cloud AI Certifications: Generative AI Leader vs Professional ML EngineerTwo very different credentials sit at the top of Google Cloud's AI ladder. Pick the one that matches the job you…45 min · beginner
- NVIDIA-Certified Associate Credentials: NCA-GENL and NCA-AIIOTwo associate-level exams anchor NVIDIA's credential ladder: one for people who build with generative AI and LLMs, one…45 min · beginner
- The Deep Learning Institute: Free Labs, Paid Workshops and GPU-Backed CoursesDLI is NVIDIA's training arm and the cheapest way to get hands on a real GPU. Its self-paced courses run in a browser…40 min · beginner
Prompt Engineering
The fastest-growing practical AI skill: designing, testing, and optimizing the instructions that steer language models — from basic prompts to multimodal and systematic optimization.
- Prompt Design FundamentalsLearn the anatomy of an effective prompt — role, task, context, format — and why clear instructions transform model…30 min · beginner
- Chain-of-Thought PromptingAsk the model to 'think step by step' and watch accuracy jump on math, logic, and multi-step reasoning tasks.30 min · intermediate
- Few-Shot LearningTeach a model new tasks instantly by showing a handful of examples right inside your prompt — no retraining required.25 min · beginner
- Context Window ManagementThe context window is the model's working memory — learn to budget it, structure it, and work around its limits.30 min · intermediate
- Temperature and Top-p SamplingThe dials that control creativity vs. consistency — understand how LLMs pick each word and how to tune randomness for…25 min · beginner
- System Prompts DesignThe hidden instructions that define an AI's personality, rules, and boundaries — learn to write system prompts like a…25 min · beginner
- Multimodal PromptingPrompting with images, audio, and documents — techniques for getting reliable answers when your input isn't just text.30 min · intermediate
- Prompt Optimization TechniquesMove from guessing to engineering: evaluate prompts systematically, version them, and squeeze out reliable performance.35 min · intermediate
- Structured Outputs, JSON Schema & Tool SchemasMake model output machine-consumable: constrained decoding, JSON Schema design, and validation at the boundary.40 min · intermediate
- AI-Assisted Coding Workflows & Review DisciplineGet real leverage from coding assistants: context hygiene, test-first loops, and review discipline that catches…40 min · intermediate
- Prompts as Code: Templates, Versioning & A/B TestingMove prompts out of f-strings and into versioned, testable, reviewable artifacts with measured rollout.40 min · intermediate
- The 5-Part Prompt Recipe That Actually WorksA repeatable five-part structure — role, task, context, format, guardrails — that turns vague AI answers into work you…30 min · beginner
- Giving AI Context: Documents, Examples and Show Don't TellPaste the source material and show two or three examples of the output you want — the single fastest way to make AI…30 min · beginner
- Never Accept the First Draft: Iterating With AITreat the first AI answer as a rough draft from a junior colleague — then use targeted follow-ups, critique loops and…25 min · beginner
- AI for Email and Professional WritingUse AI to draft, tighten and de-risk everyday business writing — difficult replies, status updates, executive…30 min · beginner
- Summarizing Long Documents and MeetingsTurn 40-page reports and 60-minute meeting transcripts into decisions, owners and risks — with techniques that keep…35 min · beginner
- Building Presentations and Slide Content With AIUse AI for the part of deck-building that actually takes time — the narrative, the one-line-per-slide message, the…30 min · beginner
- Prompt Playbook for Product ManagersConcrete AI workflows for PM work: turning research into themes, pressure-testing PRDs, drafting release notes, and…40 min · intermediate
- Prompt Playbook for QA and TestersGenerate test cases and edge cases from requirements, write bug reports developers act on immediately, and use AI to…40 min · intermediate
- Prompt Playbook for Business AnalystsTurn messy stakeholder conversations into clean requirements, user stories and process maps — and use AI to find the…40 min · intermediate
- Prompt Playbook for HR and RecruitingUse AI for job descriptions, structured interview kits, onboarding plans and policy drafts — while staying on the…35 min · intermediate
- Prompt Playbook for Marketing and ContentScale content production without sounding like everyone else: brand voice guides, repurposing pipelines, campaign…35 min · intermediate
- Prompt Playbook for Customer SupportUse AI to draft accurate replies from your own knowledge base, triage and tag tickets, spot emerging issues early, and…35 min · intermediate
- The Anatomy of a Great Prompt: Role, Task, Context, Format, ConstraintsEvery high-quality prompt is built from five interlocking parts — learn to name and assemble them so your first draft…25 min · beginner
- Why Vague Prompts FailVague prompts do not fail randomly — they fail predictably, because the model quietly fills every gap you leave with…20 min · beginner
- Role and Persona: Telling the AI Who to BeOne sentence assigning a role reshapes tone, vocabulary, and priorities before the task instruction even begins — the…25 min · beginner
- Supplying Context and Source MaterialThe model only knows what is in its training data and whatever you paste in — supplying the right source material is…30 min · intermediate
- Specifying Output FormatIdentical facts land very differently depending on shape — stating the format up front turns a pile of correct…25 min · intermediate
- Constraints, Length and ScopeConstraints are the guardrails that turn 'an answer' into 'the answer that fits where it needs to go' — set them…25 min · intermediate
- Few-Shot Examples: Show, Don't Just TellSome patterns are hard to describe in words but trivial to demonstrate — pasting two or three example input-output…30 min · intermediate
- Chain-of-Thought: When It Helps and When It Wastes TokensAsking a model to reason step by step measurably improves accuracy on hard problems — and adds pure cost with no…35 min · intermediate
- Iterative Refinement and Follow-UpsAlmost nobody nails the perfect prompt on the first try — a short first prompt followed by two or three sharp…25 min · intermediate
- Prompt Patterns Library: Summarize, Extract, Classify, Rewrite, Critique, BrainstormMost day-to-day AI requests fall into six repeatable patterns — recognizing which one you are in tells you exactly…35 min · advanced
- Testing and Comparing Prompts ObjectivelyA single impressive answer is a weak test — comparing prompts objectively means running them against a fixed checklist…40 min · advanced
- Prompt Anti-Patterns and FixesA handful of habits reliably produce weak AI output — the kitchen-sink prompt, the mind-reader prompt, the…30 min · advanced
- Bedrock Prompt Management, Prompt Optimization and FlowsStop pasting prompts into source files — version them as Bedrock resources, let the service rewrite them for a target…40 min · intermediate
- Agent Studio and the Gemini API: From Console Prompt to Production CodeIterate in Agent Studio (formerly Vertex AI Studio), then move the same prompt into the google-genai SDK — and…1.2 hours · beginner
Model Architectures
Look under the hood of GPT, BERT, diffusion models, GANs, and multimodal systems — and learn the efficiency tricks (LoRA, quantization, MoE) that make them practical.
- Transformer ModelsEncoder, decoder, or both? Map the transformer family tree — BERT, GPT, T5 — and learn which shape fits which job.45 min · intermediate
- GPT ArchitectureWalk through GPT from tokens to logits — and follow the journey from pretraining to the aligned assistant in your…45 min · intermediate
- BERT and Encoder ModelsThe bidirectional workhorse behind search engines and classifiers — how masked language modeling builds deep…40 min · intermediate
- Diffusion ModelsHow Stable Diffusion, DALL-E, and Midjourney create images from pure noise — by learning to run destruction in reverse.50 min · advanced
- GANs (Generative Adversarial Networks)Two networks locked in a forger-vs-detective duel — the architecture that first made AI-generated faces photorealistic.40 min · intermediate
- VAEs (Variational Autoencoders)Compress data into a smooth 'concept space' you can sample and interpolate — the quiet engine inside Stable Diffusion.40 min · advanced
- Multimodal ArchitecturesHow models like CLIP, GPT-4o, and Gemini connect vision, text, and audio into one shared understanding.45 min · advanced
- Efficient Model DesignsLoRA, quantization, distillation, and mixture-of-experts — the tricks that shrink giant models onto real-world budgets.45 min · advanced
- Prompting vs. RAG vs. Fine-Tuning: A Decision FrameworkA defensible way to choose between better prompts, retrieval, and training — with cost, latency, and maintenance…35 min · intermediate
- Local & Open-Weight Models: Ollama, llama.cpp and When to Self-HostRun open-weight models on your own hardware, understand quantisation formats and VRAM math, and decide when…45 min · intermediate
- Practical Fine-Tuning with LoRA and PEFTRun a real parameter-efficient fine-tune: dataset design, LoRA hyperparameters, QLoRA, evaluation, and adapter…1 hours · advanced
- Agent Architectures & Tool-Calling LoopsBuild reliable agents: the ReAct loop, state and memory, termination conditions, error recovery, and when a workflow…1 hours · advanced
- Multi-Agent Orchestration PatternsCoordinate multiple specialised agents: orchestrator-worker, pipelines, critic loops, shared state, and knowing when…1 hours · advanced
- Amazon Bedrock: One API in Front of Many Foundation ModelsThe managed gateway to 100+ models from Amazon, Anthropic, OpenAI, DeepSeek, xAI and others — plus which of its…45 min · beginner
- Amazon Nova: AWS's Own Foundation Model FamilyAmazon's first-party models — understanding, generation and multimodal — and when the price-performance case for Nova…35 min · beginner
- RAG on AWS: Amazon Bedrock Knowledge Bases, Managed and Customer-ManagedGround a model in your own documents without hand-building an ingestion pipeline — and understand the fork in the road…1 hours · intermediate
- Customising Models on AWS: Fine-Tuning, Distillation and Custom Model ImportThe four ways to change a model's behaviour on AWS, what each actually costs, and the honest test for whether you need…1 hours · advanced
- Microsoft Foundry: What the Rebrand Actually ChangedAzure AI Studio became Azure AI Foundry, which became Microsoft Foundry. Underneath the names sits a real…45 min · beginner
- Foundry Models and Azure OpenAI: Choosing, Deploying and CallingA catalogue Microsoft describes as over 10,000 models, three deployment types, two auth methods and one endpoint…55 min · intermediate
- Grounding with Azure AI Search: Vector, Hybrid, Semantic and Agentic RetrievalAzure AI Search is the retrieval engine behind almost every serious Azure RAG system. Learn the four retrieval modes…1.1 hours · intermediate
- Model Garden and Choosing a Model on EvidenceHundreds of models — Gemini, Imagen, Veo, Gemma, Claude, Llama and more — behind one platform. Learn the deployment…1.1 hours · intermediate
- Gemma: Google's Open Models and When Self-Hosting Actually WinsGemma is the open-weight family built from the same research as Gemini. Learn where it runs on Google Cloud, and the…1.2 hours · intermediate
- CUDA and the GPU Programming Model: Threads, Blocks and MemoryYou will rarely write CUDA kernels professionally, but every performance decision above you is shaped by this model…1.8 hours · intermediate
- NeMo and Multi-GPU Training: Parallelism, NCCL and Megatron TechniquesNeMo is NVIDIA's framework for customising and training models at scale, built on Megatron-style parallelism and NCCL…2 hours · advanced
- Physical AI: Omniverse, Isaac and Cosmos at Survey LevelNVIDIA's second AI business is machines that act in the world. Omniverse simulates, Isaac trains and deploys robot…1 hours · intermediate
Deployment & Scaling
Turn models into products: serving strategies, API design, cloud infrastructure, optimization, monitoring, cost control, security, and patterns for scale.
- Model Deployment StrategiesFrom notebook to production: API vs. self-hosting, batch vs. real-time, and the rollout patterns that prevent 3am…40 min · intermediate
- API Design for AI SystemsDesign APIs that tame non-deterministic, slow, occasionally-failing AI models into dependable products.35 min · intermediate
- Cloud InfrastructureGPUs, serverless, and vector stores — the cloud building blocks for AI workloads and how to choose among them.40 min · intermediate
- Model OptimizationMake inference faster and cheaper: quantization, KV caching, batching, and compilation for production serving.45 min · advanced
- Monitoring AI SystemsModels fail silently — learn to watch quality, drift, cost, and safety in production, not just uptime.35 min · intermediate
- Cost OptimizationToken bills compound fast — master model routing, caching, and prompt slimming to cut AI costs 10x without hurting…35 min · intermediate
- Security ConsiderationsPrompt injection, data leakage, and jailbreaks — the new attack surface AI applications expose and how to defend it.40 min · advanced
- Scalability PatternsFrom 10 users to 10 million: load balancing, queues, batching, and multi-GPU serving patterns for AI at scale.40 min · advanced
- Production LLM API Calls: SDKs, Streaming & RetriesTurn a demo-quality API call into a production client: official SDKs, streaming, timeouts, backoff, and idempotency.45 min · intermediate
- Token Accounting, Context Budgets & Cost ControlInstrument tokens like you instrument CPU: measure per request, budget the context, then apply caching, routing, and…35 min · intermediate
- MCP & Tool Servers: Giving Models Safe CapabilitiesExpose your systems to AI clients through the Model Context Protocol — tools, resources, transports, and permission…45 min · intermediate
- Model Serving & Inference OptimizationSqueeze throughput and latency from GPU inference: KV cache, continuous batching, quantisation, and speculative…55 min · advanced
- Building a RAG Pipeline End to EndWire the full retrieval-augmented generation path: ingestion, indexing, retrieval, grounded generation, and citations.1 hours · intermediate
- Chunking Strategies & Retrieval Quality TuningFix bad retrieval at the source: chunk boundaries, overlap, metadata enrichment, and measuring recall before touching…50 min · advanced
- Hybrid Search & RerankingCombine keyword and vector retrieval, fuse the results, then rerank with a cross-encoder for precision at the top.50 min · advanced
- LLMOps: CI/CD, Rollouts & Regression GatesShip AI changes safely: eval suites in CI, versioned prompts and datasets, canary rollouts, and rollback plans.55 min · advanced
- Observability & Tracing for AI ApplicationsInstrument LLM systems end to end: spans for every model and tool call, quality signals in production, and useful…45 min · intermediate
- Automating Repetitive Work With No-Code AI ToolsConnect the apps you already use with Zapier or Power Automate and add an AI step — so routine triage, summarising and…45 min · intermediate
- Using AI Assistants Inside Microsoft 365 and Google WorkspaceGet real value from the AI already built into the tools you use all day — where it sees your files, what it cannot do…40 min · intermediate
- Measuring Your AI Gains and Building a Team AI HabitProve the value with a simple before-and-after log, then turn individual tricks into a team capability through shared…40 min · intermediate
- Your AWS AI Workspace: Regions, IAM, Model Access and a Budget That Saves YouThe unglamorous 30 minutes that stop a learning account from turning into a surprise invoice — and stop your first…35 min · beginner
- Vector Storage on AWS: S3 Vectors, OpenSearch Serverless and Aurora pgvectorThree genuinely different answers to 'where do the embeddings live', separated by an order of magnitude in cost and…50 min · intermediate
- Amazon SageMaker AI and Unified Studio: When Managed APIs Are Not EnoughTrain, tune and host your own models — and understand the December 2024 rename that splits SageMaker the platform from…55 min · intermediate
- Strands Agents and MCP on AWS: Writing the Agent Before You Deploy ItAWS's open-source agent SDK plus the Model Context Protocol — how to give a model tools in a few dozen lines, and why…55 min · intermediate
- Amazon Bedrock AgentCore: Running Agents in ProductionThe modular platform that turns a working agent script into something you can operate — runtime, memory, gateway…1.2 hours · advanced
- Evaluating and Observing GenAI on AWS: Bedrock Evaluations and CloudWatchReplace 'it seems better' with a number — automatic evaluations, LLM-as-a-judge, human review workforces, RAG-specific…50 min · advanced
- Bedrock Cost and Scale: Caching, Batch, Throughput and Cross-Region InferenceFive levers that change generative AI unit economics on AWS by an order of magnitude — and how to attribute the spend…45 min · advanced
- Amazon Q Developer: AWS's Assistant for Building and Operating on AWSThe AWS-grounded coding and operations assistant — what it is genuinely good at, its pricing tiers, and the…35 min · beginner
- Build It: A Governed Bedrock RAG Assistant Over Your Own DocumentsAssemble everything — knowledge base, vector store, guardrail, evaluation, tracing and a cost budget — into one…1.5 hours · advanced
- Setting Up an Azure Subscription You Can Experiment In SafelySubscriptions, resource groups, regions, quota and budget alerts - the unglamorous setup that decides whether your…40 min · beginner
- Foundry Agent Service: Prompt Agents, Hosted Agents and ToolsAzure's managed runtime for agents - what the service gives you for free, the difference between a declarative agent…55 min · intermediate
- Microsoft Agent Framework: Multi-Agent Orchestration in CodeThe successor to Semantic Kernel and AutoGen. When declarative agents stop being enough, this is the open-source…55 min · advanced
- Evaluation, Prompt Flow and Observability on FoundryHow to know whether your Azure AI application is any good - golden sets, built-in evaluators, tracing, and continuous…55 min · advanced
- Azure Machine Learning: When You Still Need a Full ML PlatformFoundry did not replace Azure ML. Learn where custom models, training pipelines, MLflow tracking and managed endpoints…55 min · intermediate
- Securing Azure AI: Entra Identity, RBAC, Private Networking and AI Landing ZonesHow to make an Azure AI workload survive a security review - managed identity instead of keys, RBAC scoped properly…55 min · advanced
- Cost and Scale on Azure AI: Quota, PTU, Batch and MonitoringWhere the money actually goes on an Azure AI workload, how provisioned throughput changes the economics at scale, and…45 min · intermediate
- Copilot Studio and Microsoft 365 Copilot: Where Low-Code FitsNot every Azure AI solution should be code. Learn what Copilot Studio does well, where it hits a wall, and how it…40 min · beginner
- Opening a Google Cloud Project for AI Work Without a Surprise BillProjects, billing accounts, budget alerts, quotas, service accounts and Application Default Credentials — the…55 min · beginner
- The Gemini Enterprise Agent Platform (formerly Vertex AI): Reading the Current MapGoogle consolidated Vertex AI under the Gemini Enterprise Agent Platform name in 2026. Learn the current surface…50 min · beginner
- Embeddings and Vector Search on Google CloudTurn text, images and documents into vectors with Google's embedding models, then serve nearest-neighbour retrieval at…1.3 hours · intermediate
- Grounding and RAG: Agent Search, RAG Engine and Grounding with Google SearchFour different ways to stop a model inventing facts, from a one-line grounding flag to a fully custom retrieval…1.4 hours · intermediate
- Document AI: Deterministic Extraction Before You Reach for a ModelProcessors, OCR, form parsing, specialised invoice and identity extractors, Custom Extractor and the Layout Parser…1.2 hours · intermediate
- Function Calling, Structured Output and ToolsGive Gemini declared tools it can ask you to run, force schema-valid JSON, and understand the control-flow rule that…1.3 hours · intermediate
- Agents on Google Cloud: ADK, Agent Runtime and Agent2AgentThe Agent Development Kit for writing agents in code, Agent Runtime (formerly Agent Engine) for hosting them with…1.6 hours · advanced
- BigQuery ML: AI Where the Data Already LivesTrain classical models and call Gemini from SQL, without moving a byte out of the warehouse — plus the slot-cost…1.3 hours · intermediate
- Evaluation: Agent Platform Evals and Measuring Agents HonestlyComputation-based metrics, model-based autoraters, agent trajectory evaluation and the golden-set discipline that…1.4 hours · advanced
- Cost and Scale: Pricing, Context Caching, Batch and Provisioned ThroughputWhat actually bills on Google Cloud AI, the idle resources that quietly drain budgets, and the four levers — model…1.3 hours · advanced
- The NVIDIA AI Stack: What Sits Between Your Model and the SiliconNVIDIA is not a cloud. It is the layer every cloud runs on. Before learning any single product, learn the six-storey…50 min · beginner
- Getting GPU Access Without Buying a GPUNothing in this track requires you to own hardware. Set up three access paths first — hosted NVIDIA endpoints, a free…55 min · beginner
- GPU Selection and Memory Sizing: Choosing the Right Card on PurposeMost GPU purchases and rentals are decided by rumour. Learn to compute the memory a workload needs, read bandwidth and…1.2 hours · intermediate
- CUDA-X Libraries and Nsight: cuDNN, Compatibility, and Finding the Real BottleneckBetween CUDA and your framework sits a library layer — cuDNN, cuBLAS, NCCL — plus the profiling tools that tell you…1.5 hours · intermediate
- RAPIDS: GPU-Accelerated Dataframes and Classic MLMost enterprise AI time is spent on tabular data, not transformers. RAPIDS puts pandas, scikit-learn, NetworkX and…1.3 hours · intermediate
- NVIDIA NIM: Models Packaged as Deployable MicroservicesNIM is a container that bundles an optimised inference engine, a server, and an OpenAI-compatible API for a specific…1.4 hours · intermediate
- TensorRT and TensorRT-LLM: Compiling Models for InferenceTensorRT compiles a trained network into a hardware-specific engine — fusing kernels, selecting algorithms and…1.8 hours · advanced
- Triton Inference Server and Dynamo: Serving Many Models WellTriton is the general-purpose inference server: many frameworks, many models, one endpoint, with dynamic batching…1.7 hours · advanced
- Quantisation and Inference Economics: FP8, FP4 and Cost per Million TokensLower precision is the cheapest large win in GPU inference — but only if you measure quality alongside speed. Learn…1.6 hours · advanced
- DGX, Grace Hopper, Blackwell and Running NVIDIA AI on the CloudsA survey of the platform layer: what NVLink and NVSwitch actually change, what DGX and rack-scale GB200 systems are…1.3 hours · intermediate
Ethics & Safety
Build AI worth trusting: detect bias, engineer fairness, preserve privacy, moderate content, document transparently, and navigate accountability and regulation.
- AI Bias DetectionBias hides in data, labels, and metrics — learn where it comes from and how to measure it before it ships.35 min · intermediate
- Fairness in ML SystemsFairness has competing mathematical definitions that can't all hold at once — learn the main ones and how to choose.35 min · intermediate
- Privacy PreservationModels can memorize and leak personal data — learn differential privacy, federated learning, and practical data hygiene.40 min · advanced
- Content ModerationGuardrails for inputs and outputs: moderation APIs, policy design, and the hard tradeoffs of filtering at scale.30 min · intermediate
- Transparency RequirementsModel cards, datasheets, and explainability — documenting what AI systems do, for whom, and where they fail.30 min · intermediate
- Accountability FrameworksWho answers when AI fails? Governance structures, audits, and the NIST AI RMF that turn responsibility into process.30 min · intermediate
- Regulatory ComplianceThe EU AI Act, GDPR, and sector rules — a practical map of the legal landscape every AI builder now operates in.30 min · intermediate
- Responsible AI DevelopmentPull it all together: a practical lifecycle playbook for building AI that is safe, fair, documented, and worth trusting.30 min · beginner
- LLM Security for Engineers: Injection, Exfiltration & SandboxingThreat-model AI features: prompt injection, indirect injection through retrieved content, data exfiltration, and tool…55 min · advanced
- What NOT to Put Into AI Tools: Data, Privacy and PolicyA practical rule set for what is safe to paste, what needs an approved tool, and what should never go near an AI…30 min · beginner
- Bias, Fairness and Ethics for Everyday AI UsersYou do not build models, but you make decisions with their output. Learn where bias shows up in ordinary office use…35 min · intermediate
- Amazon Bedrock Guardrails: Filters, PII, Grounding and Automated ReasoningConfigurable safeguards that sit between your users and any model — including a formal-methods check that can prove a…45 min · intermediate
- AI Governance on AWS: IAM, Data Boundaries, Service Cards and CAF-AIThe controls and artefacts that get a generative AI workload through security review — and the organisational…45 min · intermediate
- Azure AI Content Safety: Filters, Prompt Shields and Groundedness DetectionThe platform's safety layer - what the default filters actually block, how Prompt Shields defends against injection…45 min · intermediate
- Responsible AI on Azure: Transparency Notes, the RAI Dashboard and Regulatory MappingMicrosoft's responsible AI machinery is unusually concrete - transparency notes per service, an impact assessment…45 min · intermediate
- Responsible AI on Google Cloud: Safety Filters, Model Cards and ReviewConfigurable harm-category filters, system instructions as a safety surface, model cards, and the governance process…1.2 hours · intermediate
- Securing AI Workloads: IAM, VPC Service Controls, Model Armor and Data GovernanceThe controls that get a generative AI workload past a security review: least-privilege IAM, a VPC Service Controls…1.4 hours · advanced
- NeMo Guardrails and the Agent Toolkit: Programmable Safety and Governed AgentsNeMo Guardrails puts a programmable policy layer between users, models and tools — topical, safety, execution and…1.5 hours · intermediate
Model Context Protocol
Connecting assistants to tools: the protocol, the server, the deployment, and publishing what you build.
- What MCP actually isOne protocol so any assistant can reach any tool, without an integration per pair. Plus the part most tutorials get…18 min · beginner
- Build an MCP serverDeclare a tool, validate the transport contract, return a result. You will write the manifest yourself and the checker…35 min · intermediate
- Deploy and publish an MCP serverStateless means it scales without a session store. Ship it behind HTTPS, lock the Origin check, then submit it to the…30 min · intermediate
- MCP in production: auth, gateways and blast radiusWho may call which tool, how a gateway enforces that without reading your bodies, and how to design tools whose worst…40 min · advanced