For individuals
IT Training
Instructor-led training in the engineering skills employers hire for, taught by engineers still building the systems it draws from.
Service detailTraining
AI & GenAI Training
There is a large gap between using an AI tool and engineering an AI system. This programme is about the second one: retrieval pipelines that return the right context, evaluations that catch regressions before users do, and the operational discipline to run all of it in production without surprises.
Book an admissions callFor individuals
- Format
- Live online, evenings & weekends
- Duration
- 14 weeks
- Prerequisite
- Working Python or equivalent
- Cohort size
- Capped at 14
What's included
What the programme covers
Foundations that hold up
Embeddings, tokenisation, context windows, sampling and cost mechanics — enough of the underlying model behaviour to debug a system rather than guess at it.
Retrieval-augmented generation
Chunking strategies, hybrid search, reranking, and the failure modes each one introduces. Built end to end against a document corpus that is deliberately messy.
Evaluation and regression testing
Building an evaluation set before building the feature. Offline scoring, LLM-as-judge and its known biases, and wiring evaluations into CI so quality changes are visible.
Agents and tool use
Structured tool calling, multi-step orchestration, and the practical limits of autonomy — including when a deterministic pipeline is the better engineering answer.
Classical ML where it fits
Regression, classification and gradient boosting on tabular data. A meaningful share of "AI projects" are correctly solved this way, and knowing when is part of the skill.
Safety, cost and governance
Prompt injection, data leakage, PII handling, token economics and the documentation an enterprise review board will ask you for.
Process
How the programme runs
Written down so you can hold us to it.
Technical screen
A short exercise in Python to confirm the prerequisite. This programme moves quickly and assumes you can already write and debug code without support.
Model and data foundations
Three weeks on how these models behave, how to measure them, and how to prepare data — including the unglamorous work of cleaning a real corpus.
Build a retrieval system
Four weeks constructing a production-shaped RAG service: ingestion, indexing, retrieval, generation, caching and an evaluation harness that runs on every change.
Agents and orchestration
Three weeks extending the system with tool use and multi-step workflows, then deliberately breaking it to study the failure modes.
Capstone and review
Four weeks building your own system against a specification, presented to a review panel that asks the questions a staff engineer would ask.
Outcomes
What you leave with
- A deployed retrieval-augmented system with an evaluation suite and documented performance characteristics
- A capstone project of your own design, defended in front of practising engineers
- Working knowledge of the current model APIs and the trade-offs between them
- The ability to size, cost and scope an AI feature honestly before committing to it
- A written framework for AI risk review you can bring to your own organisation
- Alumni access to office hours and the community channel, permanently
Detail
Tools and platforms covered
Models & APIs
- Anthropic Claude
- OpenAI
- Open-weight models
- Local inference
- Structured outputs
- Tool calling
Retrieval
- pgvector
- Elasticsearch
- Hybrid search
- Reranking
- Chunking strategies
- Metadata filtering
Engineering
- Python
- FastAPI
- Async patterns
- Docker
- Streaming responses
- Caching layers
Evaluation & ops
- Golden datasets
- LLM-as-judge
- Regression suites
- Tracing
- Cost monitoring
- Guardrails
Classical ML
- scikit-learn
- XGBoost
- Feature engineering
- Cross-validation
- Model interpretability
Questions
What people ask about this service
Including the ones with answers you might not want.
How much programming do I need?
You should be able to write a Python script that reads data, transforms it and calls an API without following a tutorial. If you are not there yet, the Data Engineering track in our IT Training programme is the better starting point, and you can move across afterwards.
Does this cover prompt engineering?
It covers prompting as one component of a system — structured outputs, context construction, evaluation. It is not a prompt-tips course. Prompts change with every model release; system design and evaluation discipline do not.
Which model provider do you teach?
Several, deliberately. The engineering concerns — retrieval quality, evaluation, cost control, failure handling — transfer across providers, and being locked to one vendor is a liability in a market moving this fast.
Is this useful if I am not an engineer?
Partly. Technical leads, analysts and product managers get real value from the evaluation, cost and governance modules. The build weeks assume you are writing the code yourself, so a non-engineer will get less from roughly half the programme.
Still unresolved? Ask us directly.
Related
Services that usually accompany this one
For enterprises
Data, AI & Analytics
Data platforms, analytics and production AI systems — built on foundations trustworthy enough to make decisions from.
Service detail
Next step
Check whether this programme fits
Send us your background and we will tell you honestly whether to start here or build foundations first.