15 Lessons
Hands-On Capstone
Small Language Models (SLMs) for M&E and International Development
Install, customise, and responsibly use private, efficient, and task-specific AI models for evaluation, MEAL, humanitarian, and development workflows.
No programming experience required · Local-first learning · Human judgment remains central
Free with your EvalCommunity Premium Membership — Upgrade now →
What You Will Be Able to Do
✓
Understand where SLMs fit in M&E and development work
✓
Prepare your computer and run a model locally
✓
Create a focused local workflow for one M&E task
✓
Add approved organisational knowledge and source controls
✓
Set boundaries, tripwires, escalation, and human oversight
✓
Validate the system before using it in a controlled pilot
Why This Course?
General-purpose AI tools can be useful, but they may be too broad, too dependent on cloud services, or insufficiently grounded in your organisation’s definitions, reporting rules, and evidence standards.
Small Language Models can support local, efficient, and focused workflows—but running a model is only the beginning. This course teaches you how to turn that model into a controlled M&E system with approved knowledge, validation checks, clear human responsibilities, and evidence-based limits.
Four Principles You Will Apply
Start with the task
Select the workflow before selecting the model.
Local does not automatically mean safe
Privacy also depends on configuration, access, storage, and user behaviour.
Knowledge must be controlled
Approved, current, relevant sources are more useful than an uncontrolled document collection.
AI supports—humans decide
Professional, ethical, safeguarding, and programme decisions remain accountable human work.
Who This Course Is For
M&E and MEAL officers
Evaluators and consultants
NGO and INGO staff
Humanitarian practitioners
Programme and learning teams
Researchers and data staff
Prerequisite: Basic computer use. No programming, machine-learning, or model-engineering experience is required.
5 Lessons
Module 1 — Understanding Small Language Models
Build a practical foundation before selecting or installing a model.
Understand what makes a language model “small” and how it differs from larger systems.
Connect local, efficient AI with real organisational and programme needs.
Assess privacy, efficiency, cost, quality, knowledge, and operational trade-offs.
Review model documentation, permitted uses, restrictions, and selection evidence.
Identify suitable tasks and recognise workflows that should remain human-led.
5 Lessons
Module 2 — Installing and Using an SLM Locally
Move from theory to a working local model and your first controlled M&E workflow.
Assess your device, define requirements, and choose an appropriate starting model.
Complete a beginner-friendly installation and first local run.
Turn one realistic task into a repeatable input–instruction–output process.
Create clear roles, tasks, rules, formats, and missing-information behaviours.
Control data, storage, access, network behaviour, records, and human responsibility.
5 Lessons
Module 3 — Building and Validating Task-Specific SLM Systems
Move beyond a model and build a governed, testable, organisation-aware system.
Connect users, inputs, knowledge, instructions, validation, records, and decisions.
Prepare approved definitions, guidance, templates, and source-grounding rules.
Specify purpose, authorised users, inputs, outputs, boundaries, and human roles.
Define stop conditions, escalation, handover, and accountable review.
Build test cases, score outputs, record defects, retest, and decide readiness.
Learning by Building
Design templates
Workflow maps, model-selection records, knowledge cards, and system specifications.
Worked M&E examples
Indicators, partner reports, monitoring notes, feedback, and data-quality tasks.
Failure testing
Missing evidence, conflicts, prohibited data, out-of-scope requests, and invented claims.
Human review tools
Scorecards, defect logs, tripwire registers, handover messages, and readiness decisions.
Capstone
Final Practical Assignment
Build a Small Language Model System for an M&E Workflow
Select one suitable workflow and produce a complete evidence package showing how the system was designed, controlled, tested, corrected, and approved—or why it should not proceed.
1. Workflow suitability
Justify the task and its limits.
2. System blueprint
Document people, inputs, model, output, and records.
3. Knowledge and instruction
Prepare approved sources and the core instruction.
4. Controls and red-team tests
Challenge boundaries and tripwires.
5. Validation evidence
Run at least five normal and failure cases.
6. Readiness decision
Proceed, restrict, revise, redesign, or stop.
Course at a Glance
3Modules
15Lessons
1Capstone System
100%Self-Paced
Frequently Asked Questions
Do I need programming or machine-learning experience?
No. The lessons explain each concept before asking you to apply it. Basic confidence using a computer is sufficient.
Do I need a powerful computer?
Requirements vary by model and configuration. The course teaches you how to assess your device and select a realistic starting model rather than assuming that one setup fits everyone.
Will the model work without the internet?
Some setup and download steps require internet access. A local workflow may then operate offline, depending on the runner, model, integrations, and configuration you choose.
Can I use confidential or personal programme data?
Use only information authorised under your organisation’s policies. The course starts with fictional or non-sensitive data and teaches classification, storage, access, and tripwire controls.
Does a local SLM replace professional M&E judgment?
No. The course treats the system as a bounded support tool. Final methodological, ethical, safeguarding, programme, and communication decisions remain with accountable people.
What will I have at the end?
You will have a documented, tested SLM system design for one M&E workflow, including its knowledge pack, instructions, controls, test evidence, defects, and readiness decision.
Do more than run a local model. Build an M&E system that is focused, grounded, testable, and accountable.
Free with your EvalCommunity Premium Membership — Upgrade now
Self-paced · Instant access · Lifetime updates
Course Features
- Lectures 16
- Quiz 0
- Duration Lifetime access
- Skill level All levels
- Language English
- Students 129
- Certificate No
- Assessments Yes
- 4 Sections
- 16 Lessons
- Lifetime
- Module 1: Understanding Small Language Models (SLMs)5
- Module 2: Installing and Using an SLM Locally5
- Module 3: Building and Validating Task-Specific SLM Systems5
- Final Practical Assignment1
