Twelve builds, from first prompt to production
Each lesson is a real app set in Zambia, not a toy chatbot: answers from the Employment Code, search over past papers, a WhatsApp support agent that knows when to hand over. You leave each one with working code in your own repo.
12 lessons · 3 levels · 3 open to everyone · K850 for the rest, with weekly build sessions
Not sure where to start?
Answer two questions and say what you want to build. Jev picks your level and shows how sure it is; DeepSeek turns the first lesson into a project brief for your idea.
The path
Beginner
Your first working AI app, on the internet, answering real questions.
You can write a little Python and have never called a model from code.
- B1A price assistant for a market stallA chat assistant that answers a trader’s customers: prices, stock and delivery, in the trader’s own voice.Problem and solution designLLM and prompt strategyOpen
- B2Turn a WhatsApp order into an invoice lineA function that reads a messy order message and returns clean, validated fields ready for an invoice.LLM and prompt strategyAPIs and backendLocked
- B3Ask questions of a mobile money statementA small app where you upload a statement CSV and ask “how much went on airtime in March?” in plain English.Architecture and tech stackAPIs and backendLocked
- B4Put it on the internet: a streaming chatYour B1 assistant behind a real API and a chat page that streams, recovers from errors and keeps the key on the server.APIs and backendUI and frontendLocked
Intermediate
Answers from your own documents, tools the model can call, and proof it works.
You have called an LLM API and want to build past a single prompt.
- I1Ask the Employment CodeA question-answering app over the Employment Code Act, 2019 that cites the section behind every answer and refuses when the Act is silent.RAG pipelineAuth and guardrailsOpen
- I2Search Grade 12 past papersA search engine over past exam papers that finds “osmosis, paper 2, 2019” as well as “the question about the potato strips”.RAG pipelineEvaluation and testingLocked
- I3A quotation assistant that uses toolsAn assistant that builds a quotation by calling your code: look up the price list, convert currency, add 16% VAT, draft the quote.AI agentsAPIs and backendLocked
- I4Prove it works: a test set for your RAG appAn evaluation suite for the I1 Employment Code app that scores retrieval and answers separately, and runs on every change.Evaluation and testingHuman in the loop and deploymentLocked
Advanced
Agents that take actions, guardrails, cost in kwacha, and running it in production.
You have shipped something with an LLM and want it to survive real users.
- A1A support agent that knows when to hand overA customer-support agent for a solar installer: answers from the manuals, remembers the customer, logs fault tickets, and passes refunds to a person.AI agentsHuman in the loop and deploymentAuth and guardrailsOpen
- A2Guardrails for an app that touches moneyA layered defence for the A1 agent: prompt-injection checks, personal-data filters, role checks on every tool, and a red-team test suite.Auth and guardrailsEvaluation and testingLocked
- A3Cost: paying in dollars, earning in kwachaCost tracking for your app that shows what each conversation costs in kwacha, then cuts it with caching and model routing.CostArchitecture and tech stackLocked
- A4Deploy it, watch it, keep it runningYour capstone in production: containerised, deployed by CI with an eval gate, traced, and ready to show an employer.Human in the loop and deploymentEvaluation and testingCostLocked
Learn it with people, not alone
The lessons are the path. The K850 practice is the group that walks it with you: a small group, weekly build sessions and code review, in Lusaka and remote, and a certificate with a project repo you can show employers.
Built on open work
Every lesson links to the open-source material it builds on. We set the projects, the order and the Zambian context; the credit for the underlying lessons belongs to their authors.
- AI Engineering from ScratchMIT
- Awesome LLM AppsApache-2.0
- AI Engineering Interview Questions, company-wiseApache-2.0
Levels: Beginner, Intermediate, Advanced. Interview questions are quoted with the companies that source lists; we have not checked them with those companies.