Practical AI for Developers

Course overview

Build AI features that actually hold up

For developers and engineers who can already ship software and now need to build AI-powered features that are reliable, grounded, secure and testable — not just impressive demos.

In three focused hours we go from the LLM mental model to a working approach for building AI features: calling model APIs, getting reliable structured outputs, grounding answers in your own data with RAG, using agents and tools sensibly, adding guardrails, and putting lightweight evaluation and monitoring around it. You'll leave with reusable checklists and a starter blueprint you can apply directly to your own product.

Learning outcomes

What you'll be able to do afterwards

Explain how LLMs, embeddings, RAG and agents work in practical developer terms.
Call LLM APIs effectively and get reliable, structured (JSON) outputs.
Treat prompts as versioned, testable code — not casual text.
Sketch a RAG design that grounds AI answers in your own data with citations.
Decide when to use agents and tools versus deterministic workflows.
Apply core guardrails: input/output validation, secrets, access and prompt-injection defences.
Wrap a lightweight evaluation and monitoring plan around an AI feature.
Leave with a reusable starter blueprint for building an AI feature safely.

The 3-hour agenda

7 focused modules, each with a takeaway

A single live 3-hour session (with a short break). Every module builds toward a starter blueprint you can reuse.

#	Time	Topic	You'll leave with
1	0:00–0:20	The AI developer mental model	A clear picture of LLMs, tokens, context and limits
2	0:20–0:55	Building with LLM APIs	Reliable calls and structured JSON outputs
3	0:55–1:30	RAG essentials — ground AI in your data	A RAG design sketch for your own docs
☕ Short break · 1:30–1:40
4	1:40–2:05	Agents, tools & function calling	A rule for agents vs deterministic workflows
5	2:05–2:30	Security, guardrails & safe AI coding	A guardrails & secrets checklist
6	2:30–2:50	Evaluation, testing & observability	A lightweight eval & monitoring plan
7	2:50–3:00	Recap, starter blueprint & Q&A	A blueprint you can build on next week

Detailed outline

What's inside each module

1 The AI Developer Mental Model ▾

What actually changes when AI becomes part of your app.
Traditional software vs AI-assisted vs AI-native features.
How LLMs behave: tokens, context windows, temperature and non-determinism.
What LLMs are good and bad at — and where hallucinations come from.
Common failure modes: great demo, weak grounding, no evaluation, runaway cost.

2 Building with LLM APIs ▾

Calling model APIs from your code — the essential request/response loop.
System, developer and user prompts — and why the separation matters.
Structured outputs with JSON schemas and validation you can trust.
Prompts as versioned, testable code, not casual text.
Model selection: capability, cost, speed, context size and data policy.
A thin model-abstraction layer to avoid vendor lock-in.

3 RAG Essentials — Ground AI in Your Data ▾

Why RAG matters and when it's the right pattern (and when it isn't).
Embeddings, vector search and semantic grounding in plain terms.
Ingestion: parsing, chunking, embedding and indexing your documents.
Retrieval: query rewriting, search, ranking and context assembly.
Citations and source traceability so answers stay trustworthy.
Quick wins for retrieval quality without over-engineering.

4 Agents, Tools & Function Calling ▾

Chatbot vs assistant vs workflow vs agent — the real differences.
Tool calling / function calling and how to wire it safely.
When to use an agent, when to use a deterministic workflow, when to avoid autonomy.
Safe tool access: read-only tools, approval gates and action boundaries.
Human-in-the-loop design and sensible failure handling.

5 Security, Guardrails & Safe AI Coding ▾

Input and output validation around model calls.
Prompt injection and data-exfiltration risks — and practical defences.
Secrets, keys and never exposing sensitive data to the model.
Access control and keeping data boundaries between users.
Logging and auditability without leaking sensitive content.

6 Evaluation, Testing & Observability ▾

Why probabilistic systems need a different testing mindset.
Golden question sets and expected answer patterns.
Testing hallucination, citation accuracy and refusal behaviour.
Prompt and model regression testing before release.
Lightweight observability: quality, latency, cost and usage.

7 Recap, Starter Blueprint & Q&A ▾

Pull the pieces together into a reusable AI-feature blueprint.
A pragmatic checklist to take an AI feature from idea to first release.
What to learn next and how to keep quality high as you scale.
Open Q&A on your own use cases.

Pre-work & take-aways

What you bring, what you keep

Optional pre-work

One real feature or problem where AI might help.
An example document, API or dataset it would use.
Any data-sensitivity or access constraints you know about.
Your current pain points with search, summarisation or extraction.

Post-course assets

AI-feature starter blueprint & design template.
RAG design checklist.
Prompt-engineering checklist for developers.
AI security & guardrails checklist.
Lightweight evaluation & monitoring checklist.
Glossary of AI engineering terms.

Format & pricing

One focused 3-hour live session

Live online, at the weekend. Per-person pricing below — private in-house sessions for your team on request.

Saturday 25 July 2026 · 4–7pm BST £200£100 per person

Live 3-hour masterclass 50% off launch

Developers building real AI features

The full 7-module agenda — LLM APIs, RAG, agents, guardrails and evaluation — hands-on, with the starter blueprint and all checklists included.

Book your seat → Use code DEV50OFF2026 to pay £100

Private · your team Get a quote

In-house group session

Teams wanting a tailored run

Run privately for your engineers on a date that suits you, tailored to your stack and use cases. Send the form for options.

Book your seat

Join the next cohort — Sat 25 July

A live 3-hour online masterclass for developers, Saturday 25 July 2026, 4–7pm BST. £100 per person with code DEV50OFF2026 (50% off the £200 standard price). Book instantly via Stripe.

£200£100 · code DEV50OFF2026 Saturday 25 July 2026 · 4–7pm BST · live online. Enter code DEV50OFF2026 at checkout to get 50% off.

Book your seat · £100 → Secure card payment via Stripe · code DEV50OFF2026 = £100

Booking a seat? Use the “Book your seat” button. This form is just for questions, future dates or private in-house groups.