Four days, one plant, one common data set. Day 1 uses Microsoft Copilot. Days 2 to 4 use Python in Google Colab, OpenAI and LlamaIndex, with GitHub for the course files. All data and documents are synthetic training material. Nothing here is an engineering diagnosis or a maintenance authorisation.
Instructor: Mohammad Arshad Ahmed
| Day | Theme | Platform | Start here |
|---|---|---|---|
| 1 | Generative AI, prompt engineering and data understanding | Microsoft Copilot | Day_1_Copilot_and_Data_Understanding/ |
| 2 | Building generative AI applications and RAG | Colab, OpenAI, LlamaIndex, GitHub | Day_2_GenAI_and_RAG/ |
| 3 | Agentic AI, tools and intelligent decision-making (includes a decision tree as an agent tool) | Colab, OpenAI, LlamaIndex | Day_3_Agentic_AI/ |
| 4 | Capstone, governance and final demonstration | Colab, OpenAI, LlamaIndex, GitHub | Day_4_Integrated_Capstone/ |
Scenarios are adapted to IT, cybersecurity, applications and plant operations under one common journey: asset A-001, a cooling-water pump, and the plant around it.
- Read
START_HERE.md. - Day 1: open the Excel workbooks in
data/excelthrough OneDrive or SharePoint and followDay_1_Copilot_and_Data_Understanding. - Day 2 onward: open the notebooks with the links below. The first cells clone this repository and load the data.
The Colab links work once this repository is public on the main branch. The same list is in RUN_THE_NOTEBOOKS.md.
| Day | Notebook | Minutes | What you build |
|---|---|---|---|
| Day 1 (optional lab) | Data understanding in Python | 60 | No key, no install: file types, pandas, SQL, joins, documents, quality checks |
| Day 2 | Colab setup and first prompts in code | 40 | OpenAI key, Colab Secrets, structured prompts in Python |
| Day 2 | GenAI basics and RAG from scratch | 75 | Tokens, hallucination, chunk, embed, retrieve, cited answers |
| Day 2 | Governed RAG and evaluation | 60 | Approved vs draft vs superseded, SQL + RAG, evaluation set |
| Day 3 | Tool-calling agent | 75 | Agent loop with SQL, document search and forecast tools |
| Day 3 | Decision tree as an agent tool | 75 | Train and judge a tree on log windows, give it to an agent, human review queue |
| Day 3 | Skills, memory and governed agents | 75 | Skills, memory, audit trail, approval gate, supervisor and specialists |
| Day 4 | Capstone: review briefing | 90 | Evidence packet, 8-section briefing, automatic checks |
| Day 4 | The review app, in Colab | 45 | Three-tab app inside the Colab session |
| Optional | Extension 1: explore with SQL | 50 | SQL, joins, data quality, charts, text-to-SQL with a guard |
| Optional | Extension 2: anomaly and forecast | 60 | Anomaly flags, baseline, Ridge model, 7-day vibration forecast |
| Optional | Extension 3: security logs | 50 | Investigate A-001 from raw logs, detection rules and alert fatigue |
| Need | Go to |
|---|---|
| Prompt engineering guide and practice prompts | Day_1_Copilot_and_Data_Understanding/PROMPT_ENGINEERING_GUIDE.md |
| Copilot exercises E01 to E16 with check answers | resources/COPILOT_EXERCISES.md |
| Task card for your job title | resources/PARTICIPANT_ROLES_GUIDE.md |
| All 19 documents (Word and Markdown), with status and version | documents/README.md |
| Excel workbooks for Copilot | data/excel/ |
| CSV tables used by the notebooks | data/csv/ (Asset 360) and data/security/ (security layer) |
| Slides | slides/ (one deck per day) |
| Instructor run of show | resources/INSTRUCTOR_GUIDE.md |
| Problems | resources/TROUBLESHOOTING.md |
Day_1_Copilot_and_Data_Understanding/ prompt engineering guide, Day 1 plan (no code)
Day_2_GenAI_and_RAG/ 3 notebooks: Colab setup, RAG from scratch, governed RAG
Day_3_Agentic_AI/ 3 notebooks: agent, decision tree as a tool, governance
Day_4_Integrated_Capstone/ 2 notebooks, capstone brief, rubric
Optional_Colab_Extensions/ 3 optional notebooks: SQL, forecast, security-log investigation
colab/enec_colab.py shared helper: OpenAI client, read-only SQL, RAG index, agent loop, checks
data/csv/ Asset 360 slice (12 tables)
data/security/ security layer: logins, firewall flows, controller commands, alerts, vulnerabilities, access review, windows
data/excel/ Excel Tables for Copilot
documents/ 19 documents as Word and Markdown, plus the index
skills/ agent skill playbooks
evaluation/ RAG evaluation set, security questions, instructor answer key
resources/ guides, exercises, troubleshooting
slides/ 4 decks (the original Colab-first Day 1 deck is in slides/archive)
tools/ tests/ notebook builders and the offline test
- Day 1 is Copilot-first. The earlier Colab-first Day 1 notebooks are now Day 2 setup and the three optional extensions.
- Notebooks pass an offline test with a fake OpenAI client. Run each once with a real key before teaching.
- Day 2 and Day 3 notebooks currently build RAG and agents directly on the OpenAI API so every step is visible. The Day 3 decision-tree notebook also shows a LlamaIndex tool and agent. Moving the Day 2 RAG notebooks fully onto LlamaIndex is planned.