Professional Diploma in Agentic AI & LLM Engineering

Course overview

Modern agent development requires more than prompting. Current agent frameworks combine models with instructions, tools, handoffs, guardrails, state, tracing and human approvals. This diploma covers the full stack: from LLM foundations and Python, through APIs, function calling, RAG and memory, to multi-agent systems, MCP, governance, evaluation and production deployment. It is vendor-aware but vendor-neutral: you learn transferable architecture across OpenAI, Anthropic and Google patterns rather than dependence on one product.

Who the course is for

Developers and software engineers moving into AI engineering.
DevOps and cloud engineers adding LLM systems to their skill set.
Technical graduates building an AI portfolio and career.
Technology professionals tasked with building internal AI capability.

What you will be able to do

Build secure LLM applications against production APIs with streaming, cost control and observability.
Implement function calling, structured outputs and tool permissions with confirmation before sensitive actions.
Design retrieval pipelines with embeddings, reranking, citations and access control.
Build single- and multi-agent systems with handoffs, memory, guardrails and human-in-the-loop approvals.
Develop MCP servers that connect agents to approved organisational services.
Create evaluation suites, traces and monitoring, then deploy to a cloud environment with rollback strategies.
Apply Nigeria Data Protection Act requirements and AI risk assessment to every build.

Curriculum: 16 modules

Module 1 · AI, generative AI & LLM foundations

AI/ML/deep learning, foundation models, tokens and context, capabilities and limitations, hallucination and bias, closed vs open-weight vs hosted models, model selection for cost, speed, reasoning and security, Nigerian and African AI opportunities.

Lab: Compare multiple models on a defined business task.

Module 2 · Python, Git, JSON & web foundations

Python syntax and OOP basics, virtual environments, Git and GitHub, HTTP, JSON schemas, environment variables, secure API-credential management, command line and debugging.

Lab: Build a Python application that consumes a public API.

Module 3 · Prompt engineering & LLM skill sets

System/developer/user instructions, role and context design, few-shot prompting, templates, constraint and rubric prompting, chaining, reasoning decomposition, multimodal prompting, versioning and testing, defensive prompting. Evaluation-first: define success criteria before optimising prompts.

Lab: Create and evaluate a reusable prompt library for an organisation.

Module 4 · LLM APIs

Keys and authentication, SDKs, text and multimodal requests, streaming, token usage and cost, rate limits and retries, conversation state, async, webhooks, batch, logging and observability across OpenAI, Anthropic and Gemini API patterns.

Lab: Build a secure LLM-powered web or command-line application.

Module 5 · Function calling, tools & structured outputs

Tool definitions, JSON Schema, parameter validation, structured outputs, tool selection, parallel and sequential calls, databases and external APIs, error recovery, tool permissions, confirmation before sensitive actions, idempotency.

Lab: Build an assistant that searches customer records, calculates an outcome and creates a structured report.

Module 6 · Embeddings, vector databases & RAG

Embeddings and semantic similarity, chunking, metadata design, vector stores, retrieval pipelines, hybrid search, reranking, grounded generation, citations, access-controlled retrieval, RAG evaluation and knowledge-base maintenance.

Lab: Build a question-answering assistant grounded in policy documents or company procedures.

Module 7 · Agentic AI foundations

What makes a workflow agentic; goals, plans, actions and observations; ReAct patterns; tool-using agents; planner/executor; deterministic workflows vs autonomous agents; agent state, failure patterns and when not to use an agent.

Lab: Build a research agent that gathers information, checks evidence and prepares a structured briefing.

Module 8 · Memory, context & knowledge graphs

Short/long-term memory, session storage, user profiles, episodic and semantic memory, summarisation, knowledge graphs, entity extraction, retrieval policies, privacy and deletion, preventing contaminated memory.

Lab: Build a persistent organisational assistant with controlled memory.

Module 9 · Multi-agent systems

Supervisor and specialist agents, agent-as-tool architecture, handoffs, sequential and parallel agents, debate/reviewer/evaluator patterns, shared vs isolated memory, conflict resolution, cost and latency control, debugging.

Lab: Build a coordinated team: researcher, analyst, compliance reviewer and report writer.

Module 10 · MCP & enterprise integrations

MCP architecture: hosts, clients and servers; tools, resources and prompt templates; building local and remote servers; connecting agents to files, databases and services; authentication, OAuth concepts, testing and secure deployment.

Lab: Develop an MCP server that connects an AI agent to an approved organisational service.

Module 11 · No-code & low-code AI automation

n8n, Make and similar platforms; webhooks and triggers; email and document automation; CRM workflows; human approval stages; classification and extraction; reporting; Google Workspace and Microsoft 365; error handling and monitoring.

Lab: Automate a complete SME process: enquiries, quotations, follow-ups and reporting.

Module 12 · Responsible AI, security & governance

Prompt injection (direct and indirect), data leakage, excessive permissions, tool misuse, guardrails, human-in-the-loop, audit trails, bias and fairness, explainability, risk assessments, vendor governance, incident response, Nigeria Data Protection Act requirements.

Lab: Conduct an AI risk assessment and design approval controls for an agent.

Module 13 · Evaluation & observability

Evaluation datasets, success criteria, accuracy/groundedness/relevance, tool-call correctness, LLM-as-judge limitations, human evaluation, regression tests, traces, latency and cost monitoring, production alerts, red-team testing.

Lab: Create an evaluation suite for your capstone agent.

Module 14 · LLMOps, deployment & scaling

FastAPI deployment, Docker, CI/CD, cloud deployment, secrets management, databases and caching, queues, authentication and RBAC, monitoring, prompt/model versioning, rollback strategies, cost optimisation, production support.

Lab: Deploy a secure agent application to a cloud environment.

Modules 15–16 · Capstone delivery

Identify a real business or public-sector problem; produce a solution architecture; build a functioning agent or LLM application; implement security and human approvals; create tests and evaluations; deploy; present a business case and technical demonstration. Example capstones: banking compliance assistant, government correspondence assistant, SME sales agent, loan-document analysis, HR onboarding agent, institutional knowledge assistant, multi-agent research and reporting system.

Outcome: A deployed, evaluated, governed AI solution plus a documented business case.

Assessment & certificate

Weekly labs, module assessments and a verified capstone. Graduates receive a verifiable digital certificate with a public verification code and a competency transcript listing the skills actually demonstrated: assessment scores, completed laboratories and capstone verification. We do not guarantee employment; we provide structured career preparation, portfolio development and employer engagement.

Frequently asked questions

Do I need to know Python before starting?

Basic Python is recommended. If you are new to programming, complete our four-week Python preparation module first; it is designed to feed directly into Module 2.

Can I pay in instalments?

Yes. A payment plan is available across the 16 weeks. Fees and instalment options are confirmed on application and published in the prospectus.

Is the programme available fully online?

Yes. Live virtual delivery mirrors the classroom cohort, with recordings, instructor support and the same labs and capstone requirements.

Which AI vendors do you teach?

The programme is vendor-aware but vendor-neutral: you will work with OpenAI, Anthropic and Google API patterns and learn transferable architecture rather than one product.

What do I graduate with?

A working capstone project, a GitHub portfolio, a documented business case, a technical presentation, an implementation report, a verifiable digital certificate and a competency transcript.