Save on courses or Exam now
About Instructors
-
rayanGuffy- Instructor Rating - Reviews - Students - CoursesNo biography available, please complete
🚀 Course Goals
- Stand up the whole lab yourself — Docker, PostgreSQL with pgvector, JupyterLab and Spark — each installed on camera with a version check and a smoke test.
- Call an LLM from Python and get back a validated object with an audit record, instead of loose text you have to trust.
- Ship a working AI workflow in six business functions: finance, HR, legal, sales and service, logistics, and operations.
- Take it to scale and to production: Spark, a partitioned Parquet lake, quality gates, approval gates, and a capstone brief you can trace to its evidence.
📘 About This Course
Most business AI projects do not fail at the model. They fail at the seam
between the model and the business: a number nobody can trace, a contract clause
quoted from memory, a forecast that leaked next week's data into last week's
training set, a reply sent to a customer that no one approved.
This course takes the opposite route. You install an Ubuntu lab on camera —
Docker, PostgreSQL with pgvector, JupyterLab, Apache Spark — and then apply one
pattern across the whole company: evidence first, code for the numbers,
the model for the language, and a person on anything high-impact.
Across sixty-two hands-on videos and twelve hours you work through six
business functions with their own synthetic datasets. Finance: budget against
actual, cash-flow scenarios, transaction anomalies. Human resources: retention
analytics and survey themes, with the fairness and privacy boundaries stated
before the code. Legal and compliance: clause extraction into structured fields
that carry the paragraph they came from, and obligation registers checked back
against the source. Sales, marketing and service: segmentation, campaign
attribution, and ticket triage that routes low-confidence cases to a human.
Logistics: demand forecasting, safety stock, and vehicle routing with OR-Tools
where every constraint is re-checked independently. Operations and quality:
KPIs with control limits, predictive maintenance, bottleneck analysis.
Then you take it to scale. Spark schemas, partitions and execution plans; a
partitioned Parquet lake; data-quality gates that block a bad month before it
reaches a dashboard; semantic retrieval on PostgreSQL and pgvector with a
citation check on every answer. The last module is governance and a three-part
capstone: secrets and least privilege, audit logs with cost and latency, human
approval gates, and an executive brief you can trace statement by statement back
to the row it came from.
Every dataset is synthetic and generated for the course. Every piece of
software you need is installed on camera, from its official source, with a
version check and a smoke test — nothing appears between cuts. Every lesson
ships with its slide deck and the code that was actually run.
This is decision support, done honestly. The model drafts and explains;
SQL, Python, Spark and OR-Tools compute; and a named person signs off before
anything leaves the building.
What You'll Learn
Stand up a complete analytics and AI lab on one Ubuntu machine: Docker, PostgreSQL with pgvector, Adminer, a Python environment, JupyterLab and Apache Spark — each installed on camera with a version check and a smoke test.
Call an LLM API from Python and get back a validated object instead of loose text, with an audit record of the model, the tokens, the latency and the validation result.
Write data contracts and choose the analytical grain before you analyse, and decide deliberately between pandas and Spark for a given job.
Build finance workflows end to end: budget against actual, variance by cost centre, cash-flow and working-capital scenarios, transaction anomaly detection, and grounded commentary for management reporting.
Apply AI inside HR and legal boundaries: retention analytics as decision support, skills taxonomies, survey themes, clause extraction with paragraph references, obligation registers and deviation analysis against a template.
Run revenue and operations analytics: segmentation, campaign attribution, ticket triage with confidence thresholds, demand forecasting, safety stock, OR-Tools routing, control limits and predictive maintenance.
Work at scale with Spark: schemas, partitions and execution plans, a partitioned Parquet lake, large joins and windows, and data-quality gates that stop a bad month from reaching a dashboard.
Govern the whole thing: secrets and least-privilege access, audit logs with cost and latency, human-in-the-loop approval gates, and a capstone executive brief whose every statement traces back to validated evidence.
🛠️ Requirements
- Comfortable on the Linux command line and with basic SQL and Python. You do not need prior machine-learning or data-engineering experience.
- A machine or VM with about 8 GB of RAM, 4 CPU cores and 30 GB of free disk, running Ubuntu 22.04 or 24.04.
- Access to an OpenAI-compatible LLM endpoint for the lessons that call a model. The course treats it as an external governed dependency and verifies its capabilities before relying on them.
- No GPU and no paid cloud account. Everything else is installed during the course, from its official source, with a version check.
👤 Who This Course is For
- Analysts and business-function specialists in finance, HR, legal, sales, logistics or operations who want to build AI workflows for their own domain rather than read about them.
- Data analysts and data scientists who can already model, but need the evidence, validation and approval scaffolding that makes output publishable.
- Engineers asked to put an LLM behind a business process and to prove, afterwards, where each number came from.
- Technical leads and managers who have to review, approve or govern AI work and want to see exactly what good looks like.
53618 Students applied
100 Comments
-
-
John DoeMorbi nec nisi ante. Quisque lacus ligula, iaculis in elit et, interdum semper quam. Fusce in interdum tortor. Ut sollicitudin lectus dolor eget imperdiet libero pulvinar sit amet
-
John Doe
Morbi nec nisi ante. Quisque lacus ligula, iaculis in elit et, interdum semper quam. Fusce in interdum tortor. Ut sollicitudin lectus dolor eget imperdiet libero pulvinar sit amet
-
-
John Doe
Morbi nec nisi ante. Quisque lacus ligula, iaculis in elit et, interdum semper quam. Fusce in interdum tortor. Ut sollicitudin lectus dolor eget imperdiet libero pulvinar sit amet
-
John Doe
Morbi nec nisi ante. Quisque lacus ligula, iaculis in elit et, interdum semper quam. Fusce in interdum tortor. Ut sollicitudin lectus dolor eget imperdiet libero pulvinar sit amet
-
John Doe
Morbi nec nisi ante. Quisque lacus ligula, iaculis in elit et, interdum semper quam. Fusce in interdum tortor. Ut sollicitudin lectus dolor eget imperdiet libero pulvinar sit amet
Morbi nec nisi ante. Quisque lacus ligula, iaculis in elit et, interdum semper quam. Fusce in interdum tortor. Ut sollicitudin lectus dolor eget imperdiet libero pulvinar sit amet