Corporate cohort
A team upskilled together
- Run against your stack, and your data where you want it
- Scheduled around working hours
- Exercises drawn from problems your team recognises
- Sessions recorded for anyone who misses one
Training
Learn data and AI from
the people who ship it.
Three courses in data science, data engineering, and AI — taught on the same stack we use for client work, by engineers who build it the rest of the week.
Where each course fits
One pipeline, three places to join it
Raw data
Logs, exports, third-party APIs, the spreadsheet someone maintains by hand.
What you start with
Pipelines & warehouse
Moved, modelled and tested on a schedule, so the numbers can be trusted.
Data Engineering
Analysis & models
Questions answered, patterns found, predictions validated before anyone acts on them.
Data Science
Shipped in the product
An AI feature real users touch, grounded in your data and measured in production.
AI & Machine Learning
The courses
Which one fits you?
Each course runs from fundamentals to production practice. Take one, or run them in sequence — engineering after science, AI after either.
Data Science
Turn a messy dataset and a vague question into an answer people can act on.
- Best if
- You already work with data in spreadsheets or SQL, and want to analyse and model it properly.
- You build
- An analysis, end to end
Data Engineering
Build pipelines and warehouses other people can run without fear.
- Best if
- You write SQL or code, and you now own — or are about to inherit — data pipelines.
- You build
- A tested pipeline
AI & Machine Learning
Ship an AI feature that holds up in front of real users, not just a demo.
- Best if
- You can already code, and you now need to put a model into a product.
- You build
- A grounded assistant
Course 01
Data Science
Turn a messy dataset and a vague question into an answer people can act on.
Best if: You already work with data in spreadsheets or SQL, and want to analyse and model it properly.
What you cover
- 01
Python for analysis
FoundationHandle real data without corrupting it.
pandasNumPyjoinsreshapingmissing data - 02
Statistics you can defend
FoundationKnow when a result is real.
distributionssamplingconfidence intervalshypothesis tests - 03
Exploratory analysis
AppliedFind the shape of the data before modelling it.
profilingoutlierscorrelationsegmentationcharting - 04
Modelling
AppliedBuild a predictive model and validate it honestly.
regressionclassificationfeature prepcross-validation - 05
Communicating results
ProductionGet a decision made from your analysis.
assumptionsuncertaintynarrativestakeholder review
What you can do after
- Clean and join real datasets without silently dropping or duplicating rows
- Pick the right test, and state honestly how confident the result is
- Build a validated model and explain what it will and will not predict
- Present an analysis that survives a senior stakeholder pushing back
You build
An analysis, end to end
Take a raw dataset through cleaning, modelling and validation to a written recommendation.
Tools
PythonpandasNumPyscikit-learnJupyterSQLMatplotlib
Course 02
Data Engineering
Build pipelines and warehouses other people can run without fear.
Best if: You write SQL or code, and you now own — or are about to inherit — data pipelines.
What you cover
- 01
SQL past the basics
FoundationWrite queries that stay fast as the data grows.
window functionsCTEsindexesEXPLAIN plans - 02
Warehouse modelling
FoundationShape tables so queries stay cheap.
dimensional modellingslowly changing dimensionspartitioning - 03
Pipelines & orchestration
AppliedMake every run safe to repeat.
ETL vs ELTidempotencybackfillsAirflow DAGsretries - 04
Transformation as code
AppliedVersion, test and review your business logic.
dbt modelstestsdocumentationCI - 05
Quality, monitoring & cost
ProductionFind out before your users do.
data contractsfreshness checksalertingwarehouse spend
What you can do after
- Design a pipeline a colleague can operate without calling you
- Debug a failed run from the logs instead of guessing
- Rerun and backfill safely, without double-counting
- Explain — and reduce — what a warehouse bill is actually paying for
You build
A tested pipeline
Ingest, transform and test a dataset on a schedule, with alerting when it breaks.
Tools
SQLAirflowdbtSparkSnowflakeBigQueryDatabricksKafka
Course 03
AI & Machine Learning
Ship an AI feature that holds up in front of real users, not just a demo.
Best if: You can already code, and you now need to put a model into a product.
What you cover
- 01
How models actually behave
FoundationKnow where they are reliable and where they invent.
tokenscontext windowstemperaturecostfailure modes - 02
Prompting & structured output
FoundationGet predictable, parseable answers.
system designJSON schemasrefusalstruncation - 03
Retrieval (RAG)
AppliedMake the model answer from your content, not its memory.
chunkingembeddingsvector searchhybrid retrievalcitations - 04
Evaluation
AppliedProve a change actually made it better.
test setsscoringregressionshuman review - 05
Shipping safely
ProductionSurvive strangers typing into your product.
rate limitsprompt injectionabuse handlingcost ceilings
What you can do after
- Ground answers in your own content so the model quotes you, not itself
- Measure whether a prompt or model change improved quality, rather than guessing
- Bound the worst-case bill before it arrives
- Handle prompt injection and abuse instead of hoping nobody tries
You build
A grounded assistant
A retrieval-backed assistant over your own documents, with an evaluation set to score it.
Tools
Pythonembeddingsvector searchRAGLLM APIseval harnesses
The impact
What should be different afterwards.
Training is only worth the time it costs if something changes. These are the shifts the courses are built to produce — stated as capabilities you can check for, not scores we would have to invent.
Before
Reports rebuilt by hand every month
After
Scheduled pipelines that run, test and alert themselves
Before
Nobody can explain why two dashboards disagree
After
Tested transformations with lineage you can point at
Before
One person is the only one who can fix the pipeline
After
Runbooks and code the whole team can operate
Before
Analysis that falls apart under a hard question
After
Stated assumptions, quantified uncertainty, a defensible answer
Before
AI features that demo well and fail with real users
After
Grounded retrieval with an evaluation set behind every change
Before
A cloud bill nobody can account for
After
Costs traced to the queries and jobs causing them
How it runs
Three ways to learn with us.
Length, schedule and depth are agreed with you before anything starts. We would rather scope it around your team than sell a fixed course half the room already knows.
Individual & small group
Someone changing direction
- Live sessions, not pre-recorded video
- Written feedback on the work you submit
- More time on each person's own code
- A finished project you keep
Workshop or review
One specific gap
- Scoped to a single topic and outcome
- Works as a follow-on after a cohort
- Can review something you have already built
- Shortest way to unblock a team
Why us
We teach it because we ship it.
Working engineers teach it
The people running sessions spend the rest of the week building this for clients.
What we buildReal systems, not toy data
Late-arriving rows, schemas that change without warning, models that flatter themselves.
Assessed on our own platform
Coding and MCQ assessments that run in the browser, with per-topic analytics.
Try the exam portalYou keep what you build
Notebooks, pipelines, project code and written material are yours at the end.
See our free toolsGet in touch
Let's build your
data solution.
From automated reporting to cloud data pipelines and custom analytics tools. No commitment required — tell us what you're working on.
Available for remote projects worldwide
Built with precision. Powered by data.