Cloud

Mastery 2026

July 16th

The Social House, Nairobi

Registration closed

You're not watching a demo. You're building one

Forget workshops where you watch someone else build things. At Cloud Mastery 2026, you build SokoAI — a fully functional AI super-agent, grounded in the Kenyan market, running in your own Google Cloud project by the end of the day. Shop inventory. Auto parts matching. Financial advisory. Customer support. One agent. Built by you, layer by layer, from 09:00 to 16:00.

Why Cloud Mastery 2026?

Immediately Relevant Scenarios: Focused entirely on Kenyan business contexts—such as grocery retail, auto parts, and personal finance—so you see the business case before writing a single line of code.
Full-Stack Architecture: Covers everything from serverless deployment and conversational AI to machine learning, showing you exactly how each layer connects and why.
Tangible Deliverables: You leave with a live, functioning AI agent deployed in your own Google Cloud project to demo and extend, rather than just notes and slides.
Genuinely Agentic Technology: Move beyond rigid chatbots; you will build a dynamic system (SokoAI) that reasons, calls tools, and constructs answers from live data.

Download Curriculum

Highlights from our past event

See full highlights

Day at a Glance

08:30 - 09:00

DOORS OPEN

Grab a coffee, get your environment online, and find your seat.

09:00 - 09:30

GOOGLE GUEST SPEAKER: Kenneth Olang, Platform Customer Engineer

The strategic case for Agentic AI - why it matters for your organisation right now.

09:30 - 10:00

KEYNOTE: The Soko AI Vision (Oscar Limoke, CTO)

Agentic AI isn't a feature. It's the new application layer. Oscar opens the day by introducing SokoAI - what it is, what you'll build, and why it changes how software gets made.

10:00 - 10:30

GRAND DEMO: Soko AI Live (Eddie Ngugi, Head of Engineering)

Before you build it, you see it. Eddie demonstrates the completed SokoAI agent live - then opens the floor to one question every engineer in the room will have: how did that actually work?

10:30 - 11:30

SESSION 1: Serverless & Secure Foundation (Evans Mwongera, Cloud Architect)

Deploy SokoAI's core on Cloud Run - serverless, production-grade, and secure from day one. Connect your app to a hosted Cloud SQL database and import the SokoAI product catalogue.

11:30 - 12:30

SESSION 2, Part 1: The Agent Brain (Eddie Ngugi, Head of Engineering)

Introduction to Google's AI stack - Vertex AI Agent Builder, Gemini, and Conversational Agents. The brain gets its first layer.

12:30 - 13:30

NETWORKING LUNCH

Three courses. Trainers at the table. You already have a deployed app - this is a good time to show it off.

13:30 - 15:00

SESSION 2, Part 2: The Agent Brain (Eddie Ngugi, Head of Engineering)

Wire SokoAI's live inventory into your conversational agents. Ground them in real product data, test the conversations, and deploy SokoAI as a web widget on your storefront. The app becomes intelligent.

15:00 - 16:30

SESSION 3: Data Intelligence (John Higi, Senior Data Engineer)

BigQuery ML - K-Means customer segmentation and demand prediction, all in SQL. Feed the model outputs back into SokoAI so the agent's recommendations get smarter with every interaction.

17:00

VIBE & CONNECT

The build is done. Tea, snacks, and open floor time with trainers and peers.

Curriculum

09:30 - 10:00 AM

KEYNOTE: The Soko AI Vision

Agentic AI isn't a feature. It's the new application layer. The keynote opens the day by framing the fundamental shift — from apps that respond to clicks, to agents that reason, plan, and act. It introduces SokoAI, the product the room will build together, and maps out the architecture from the ground up.

What you'll understand:

What 'agentic' means in practice. The three SokoAI modules — Shop, Parts, and Wealth — and the business problem each one solves.

Oscar Limoke

10:00 AM - 10:30 AM

GRAND DEMO: Soko AI Live

Before you build it, you see it. The completed SokoAI agent is demonstrated live — attendees can interact with it from their phones via QR code. Then the architecture is pulled up and walked through in full, connecting what you just saw to the code and services you'll build across the rest of the day.

What you'll understand:

The finished product, end to end. Every technical decision you make from this point has context.

Eddie Ngugi

10:30 AM - 11:30 AM

SESSION 1: Serverless & Secure Foundation

Deploy SokoAI's core infrastructure on Cloud Run — serverless, production-grade, and secure from day one. You'll containerise the Next.js frontend and NestJS API, configure Workload Identity Federation to eliminate stored credentials entirely, and connect to Cloud SQL via the Auth Proxy. By the end of this session, your SokoAI storefront is live in your browser, serving real product data from your own database.

What you'll build:

A running Next.js + NestJS application on Cloud Run, connected to a Cloud SQL instance seeded with the SokoAI product catalogue — with zero stored service account keys.

Evans Mwongera

11:30 AM - 12:30 PM

SESSION 2, Part 1: The Agent Brain

Introduction to Google's AI stack — Vertex AI Agent Builder, Gemini, and Conversational Agents. This session walks through the SokoAI multi-agent architecture: one orchestrator routing to three specialist sub-agents (Shop, Parts, Wealth), each with its own tools and data grounding. The brain gets its first layer.

What you'll understand:

How Vertex AI Agent Builder structures agents, tools, playbooks, and data stores — and how the SokoAI architecture maps to a real business use case.

Eddie Ngugi

1:30 PM - 3:00 PM

SESSION 2, Part 2: The Agent Brain

Now you build it. Wire SokoAI's live inventory into your conversational agents, ground each sub-agent in real product data, and test the conversations. Then deploy SokoAI as a web widget directly on your Cloud Run storefront. By 15:00, your app is conversational — responding to real queries from real data in real time.

What you'll build:

Three configured sub-agents (Soko Shop, Soko Parts, Soko Wealth) live in Vertex AI Agent Builder, integrated into your Next.js frontend via the Agent API with streaming responses.

Eddie Ngugi

3:00 PM - 4:30 PM

SESSION 3: Data Intelligence

Every search, click, and order SokoAI handles becomes a training signal. This session shows how Cloud Functions stream activity data into BigQuery in real time, then builds two ML models entirely in SQL — a K-Means clustering model to segment users by behaviour, and a classification model to predict conversion. The outputs feed directly back into SokoAI's recommendations, closing the loop. The agent gets smarter from its own data.

What you'll build:

A live data pipeline from Cloud SQL to BigQuery, two BigQuery ML models in SQL, and a Looker Studio dashboard showing agent activity, user segments, and conversion predictions — all connected back into SokoAI.

John Higi