Skip to content
Segun UmoruSEGUN UMORU
Cohort 1 · Applications open

Build, Deploy and Scale Production-Ready AI Systems

11 intensive weekends. From Python and LLMs to RAG, AI agents, MLOps and production deployment.

110 hours

Live contact time

6 projects

Plus a production capstone

30 seats

Cohort 1 only

Apply for a seatStarts Saturday 10 October 2026 · Early bird closes 30 September 2026
Trained & delivered for
LOTUS Bank3MTT and Data Science NigeriaDigital RegenesysGOMYCODEWAAW Foundation

Why this programme

Most AI courses stop at the API call.

Most AI courses teach you to call an LLM API. The gap between that and an AI engineer is everything that comes after the call: retrieval that returns the right document, evaluation that catches a wrong answer, deployment that survives traffic, and a cost model that does not bankrupt the feature. That gap is the entire curriculum.

Build → Test → Evaluate → Deploy → Monitor → Improve. Six times, before the capstone.

Production, not prototypes

Every module ends in something that runs. Build, test, evaluate, deploy, monitor, improve — and you run that loop six times before the capstone.

AI system design as a first-class skill

A full module on architecture: ingestion, SQL versus vector storage, retrieval, agents, queues, caching, auth, observability, evaluation, cost, latency, security, human-in-the-loop.

African business cases, not toy problems

Fintech, logistics, FMCG, customer support and research. The projects look like systems companies here actually pay to have built.

Evaluation taken seriously

Groundedness, retrieval metrics, test sets, hallucination detection, prompt injection and guardrails. Most programmes skip this. It is the difference between a demo and a system anyone trusts.

An employer-facing capstone

Architecture diagram, public repo, deployed URL, README, tests, evaluation report and a five-minute demo — presented at a public Demo Day with employers invited.

Curriculum

Eleven weekends, eleven shipped things.

Saturdays are concepts, architecture and live build-along. Sundays you build, reviewed in the room. Wednesdays are office hours.

  1. 01

    10 – 11 Oct

    AI Engineering Foundations

    Python for engineers, Git workflow, environments, FastAPI, testing.

    A tested API service, in a repository, running.

  2. 02

    17 – 18 Oct

    ML and Deep Learning Essentials

    Data pipelines, supervised learning, metrics, neural-network intuition.

    A trained model behind an API endpoint.

  3. 03

    24 – 25 Oct

    LLM Foundations

    Transformers and tokens in practical terms, model APIs and selection, prompting patterns, structured outputs, cost and latency as constraints.

    A structured-output service with cost instrumentation.

  4. 04

    31 Oct – 1 Nov

    Building LLM Applications

    OpenAI, Anthropic and open models. Streaming, conversation state, memory, retries, rate limits, graceful degradation.

    Project 1 — Fintech AI analyst over transaction data.

  5. 05

    7 – 8 Nov

    RAG Engineering

    Chunking, embeddings, vector databases, hybrid search, reranking, metadata filtering, and making sure the current document wins.

    Project 2 — Customer support RAG with citations.

  6. 06

    14 – 15 Nov

    RAG Evaluation and Reliability

    Groundedness and faithfulness. Retrieval metrics separated from generation metrics. Test sets, hallucination detection, regression testing.

    An evaluation harness and report for your Week 5 system.

  7. 07

    21 – 22 Nov

    AI Agents

    Tool calling, function schemas, planning and state, graph orchestration, the Model Context Protocol, and when an agent is the wrong answer.

    Project 3 — Logistics agent handling queries and exceptions.

  8. 08

    28 – 29 Nov

    Advanced Agent Systems

    Multi-agent patterns and when they earn their complexity. Human-in-the-loop without the bottleneck. Guardrails, approval gates, blast radius, audit trails.

    Project 4 — FMCG sales agent with an approval gate.

  9. 09

    5 – 6 Dec

    Open-Source Models and Fine-Tuning

    Hugging Face, local inference, LoRA and PEFT, quantisation, serving. When fine-tuning beats prompting and retrieval, and when it does not.

    Project 5 — a fine-tuned model serving a narrow task cheaper.

  10. 10

    12 – 13 Dec

    Production AI and MLOps

    Docker, CI/CD, cloud deployment, secrets. Observability and tracing for systems that reason. Cost control, caching, rollback.

    An earlier project deployed publicly, with monitoring.

  11. 11

    19 – 20 Dec

    AI System Design and Security

    Full architecture under constraints. Multi-tenancy, data privacy, prompt injection and defence. Capstone architecture review.

    Capstone architecture, reviewed and signed off.

Capstone build period: 21 December – 15 January, with scheduled reviews.

Demo Day is Saturday 16 January 2027 — a public showcase with employers invited. Five minutes, live, to an audience and a technical panel. It is the single most useful thing the programme produces.

Apply

Choose a tier and tell me where you are.

Applications, not instant checkout. I read every one, we have a short conversation to check the fit, and only then does an invoice follow. Nothing is owed before that.

No payment now. 30 seats · early bird closes 30 September 2026.

Before you apply

What you need

  • Comfortable writing basic Python: functions, loops, data structures.
  • A laptop with at least 8GB RAM and a stable internet connection.
  • No prior machine learning, LLM or cloud experience assumed.
  • Roughly 6–8 hours a week outside sessions for project work.

Schedule

Saturday · 10:00 – 14:00 WAT
Concepts, architecture, live build-along
Sunday · 14:00 – 18:00 WAT
Lab — you build, reviewed live
Wednesday · 18:00 – 20:00 WAT
Office hours, code review, blockers

Questions

Answered plainly

Will sessions be recorded?
Yes, and available during the programme and after. Live attendance is strongly encouraged — the Sunday labs are where most of the learning happens.
What if I miss a weekend?
Watch the recording and bring questions to Wednesday office hours. Missing more than two weekends makes the capstone difficult, and we will talk before that becomes a problem.
Do I need a strong maths background?
No. You need to be comfortable in Python. The maths is explained where it is needed and skipped where it is not.
Is there a job guarantee?
No, and be sceptical of anyone offering one. What you get is a deployed portfolio, a public Demo Day presentation, interview preparation and an alumni network.
Can my employer pay?
Yes. Invoices can be issued to a company, and private corporate cohorts are available.
What happens after the 11 weekends?
The capstone build period runs to Demo Day on 16 January, with scheduled review sessions. Community access continues afterwards.