Free · no sign-in
Getting Started with AI
8 sessions · Beginner
Eight sessions on how language models really behave: tokens and sampling, cost and latency, prompts and context, structured output, choosing a provider, and using AI responsibly.
Guided program · Architecture track
Design AI systems that can be trusted beyond the demo.
This program is for experienced software and infrastructure professionals who want to move toward Cloud/AI Solutions Architect work. It starts with the same AI/ML and agentic-tools fluency as the FDE path, then goes deeper into architecture judgment: workload shape, data, identity, cost, resilience, oversight, and the decisions a model cannot safely make on its own.
Apply / book a callArchitecture starts with fluency
Cloud experience is valuable, but it is not enough to bolt an AI label onto an old infrastructure diagram. The shared spine builds a working model of AI systems first: how models behave, how retrieval and tools change the system, and how to evaluate and secure the result.
You should already be comfortable with software and deployment concepts. Your cloud depth can vary; the program makes the AI foundation common while giving the architecture tail enough room to examine trade-offs rather than only follow a service tutorial.
Start with the free courses
The foundation of this program is public. Read it online, free and without signing in, before you decide on anything.
Free · no sign-inBuild and Evaluate RAG Applications
8 sessions · Intermediate
Eight sessions on retrieval-augmented generation: embeddings and vector search, grounded answers with citations, why retrieval fails, evaluation suites and catching hallucinations.
Free · no sign-inTaking AI to Production
9 sessions · Intermediate
Nine sessions on running AI systems for real: where models run, reference architectures, scaling, cost, observability and audit, ending in a design you defend with decision records.
Program shape
A shared spine, then architecture judgment
Plan for roughly 4–6 months part-time. The final architecture work is paced around your starting depth and the complexity of the capstone.
Stage 1
1. Build the shared AI fluency
Learn what models compute, how cost and latency affect design, and how to direct AI coding and agentic tools with context, constraints, and machine-usable output.
- What a language model actually computes
- Context engineering
- Provider abstraction and model selection
Stage 2
2. Ground systems in data
Design retrieval and grounding that can be tested, cited, and debugged rather than trusting a demo that happens to look convincing.
- Embeddings and vector search
- Retrieval-augmented generation end to end
- Why retrieval fails in production
Stage 3
3. Understand tools and agents
Reason about tool use, protocols, memory, orchestration, and the boundaries where an agent needs supervision or a deterministic workflow.
- Tool use and function calling
- Model Context Protocol
- Workflow orchestration for long-running work
Stage 4
4. Operate trustworthy AI
Make evaluation, guardrails, observability, responsible AI, and security part of the architecture instead of afterthoughts.
- Evaluation suites
- Observability for AI systems
- Security for AI systems
Stage 5
5. Design AI systems on cloud infrastructure
Diverge into workload shapes, reference architectures, scaling, data architecture, and cost decisions that you can defend to a technical and business audience.
- AI workload shapes and where they run
- Reference architectures for retrieval and agentic systems
- Cost architecture
Stage 6
6. Establish control and evidence
Design identity, approval, isolation, and auditability into the system, then present an end-to-end architecture with a clear decision record.
- Identity and access for agents
- Human-in-the-loop approval workflows
- Decision records and defending a design
What makes this architect-shaped
Choose a shape before choosing a service
The work begins with requirements and constraints. You will compare retrieval, agentic, and workflow shapes before reaching for a cloud product.
Defend trade-offs
A useful architecture explains cost, latency, resilience, data boundaries, identity, and operational ownership to more than one kind of stakeholder.
Keep humans in control
Approval workflows, isolation, auditability, and guardrails are design material. Autonomy is not the same thing as a safe system.
Leave an inspectable record
The capstone becomes a design record: assumptions, alternatives, decisions, risks, and the evidence that supports the chosen path.
Proof is being built carefully
The AI-first framing is new. We will publish alumni stories only when the underlying outcome is verified and the person has consented to the use of their name or details.
Frequently asked questions
Do I need to be a DevOps engineer?
No single job title is required. Existing experience in software, backend, QA, DevOps, or adjacent technical work is useful. Cloud depth can vary; the program will not assume that every participant has the same infrastructure background.
Is this a cloud certification course?
No. It is a guided architecture program. Cloud services and infrastructure are used as the setting for decisions about AI systems, not as a list of badges to memorise.
How long does it take?
The realistic public estimate is 4–6 months part-time. Your exact schedule depends on your starting point, available weekly time, and the scope of the capstone.
How is pricing decided?
Guidance-based pricing — talk to us. We discuss fit, background, and delivery shape before quoting rather than placing a fixed shopping-cart price on an adaptable professional program.