Training
Enterprise AI training for engineering teams. Practical, lab-led sessions take developers from a first session to confident, governed use of AI-assisted development, and finish with a working project.
Hands-on enablement for engineering teams, including Claude Code training, taught with guardrails, permissions and review built into every lab.
- Claude Code CLI
- IDE integrations
- Git and GitHub
- Skills, subagents and hooks
- Python
- FastAPI
- SQLite
- pytest and ruff
- Slide decks
- Participant handouts
- Lab guides
- Trainer guides
The challenge
Teams are handed powerful AI coding tools with little guidance. Some engineers avoid them, others use them without limits, and habits form unevenly. The same tool produces careful, verified work in one hand and unreviewed changes in another.
Enterprises then face a second problem: nobody can say what the tool is allowed to touch, who reviews its output, or how usage is tracked. Adoption stalls, or it spreads without the controls that compliance and security teams expect.
What it does
Foundations and first session
Install, sign in and run the tool from the terminal and an IDE. Participants learn how the agentic loop reads files, runs commands and edits code, and where it fits in daily work.
Context and prompting
Managing the context window, writing specific prompts with file references and clear success criteria, and recovering cleanly with clear, compact and rewind when a session drifts off course.
Project memory and settings
Creating and maintaining a project instruction file, understanding user, project and local settings, and keeping team conventions consistent so every session starts from the same shared guidance.
Plan, build, verify, commit
A disciplined workflow: explore, plan, implement and commit, with plan mode and test-based verification. Includes debugging, safe refactoring, data analysis and turning rough ideas into written specifications.
Permissions and safety
Permission modes, allow and deny rules, sandboxing, secrets handling and diff review. Participants work from a safety checklist and a settings template they can reuse in their own repositories.
Extending and automating
Building custom skills, subagents and hooks, connecting tool servers, running the tool non-interactively in pipelines, and working in parallel sessions. Ends with an individual capstone and an adoption plan.
Training programmes
- AI harness engineering training
- Generative AI training for developers
- AI agent development training
- AI agent architecture workshop
- LLM application development training
- RAG development training
- LLMOps training
- AI engineering workshops and corporate AI upskilling
How it works
Scope the audience
We agree the audience, current experience, tools in use and goals with your team, then adjust the sessions and labs to match your stack and policies.
Teach in short blocks
Each concept block is a short demo followed within minutes by a hands-on lab. Every participant works at their own machine, with stretch goals for those who finish early.
Practise on a shared repo
Labs run against a prepared practice repository with tests and linting, so each exercise ends in a check the tool can run and the participant can inspect.
Assess and plan adoption
A recap quiz, an individual capstone with lightning demos, a scorecard and feedback form close the course. Teams leave with a 30, 60 and 90 day adoption plan.
Under the hood
- Tools taught
- Claude Code CLI
- IDE integrations
- Git and GitHub
- Skills, subagents and hooks
- Practice repository
- Python
- FastAPI
- SQLite
- pytest and ruff
- Course materials
- Slide decks
- Participant handouts
- Lab guides
- Trainer guides
- Assessment
- Recap quizzes
- Capstone scorecard
- Peer demos
- Completion certificate
Business outcome
Confident, consistent daily use. Engineers use AI assistance as a normal part of their workflow, with a shared way of prompting, planning and verifying instead of individual improvisation.
Safer by default. Teams configure permissions, sandboxing and review habits from day one, reducing the chance of unreviewed or unintended changes reaching the codebase.
Reusable team assets. Participants leave with project instruction files, skills, subagents and settings templates that the whole team can adopt and maintain.
A clear adoption path. Leads receive an adoption playbook and a 30, 60 and 90 day plan so early momentum turns into a governed, repeatable practice.
Where it is used
- Engineering teams starting with AI-assisted development and wanting a safe, shared baseline before wider rollout.
- Technical leads who need review, permission and cost habits agreed before engineers get broad access.
- Teams already experimenting who want to standardise skills, hooks and settings across projects.
- Mixed groups of analysts and engineers who need data analysis, specification writing and automation skills.
Frequently asked questions
How long is the course?
The core programme is two days of eight hours each, delivered in person or live online. Shorter or extended formats, and extra topics such as extending the tool, can be agreed for your team.
Do participants need prior experience?
No. The course is written for working professionals new to the tool. Terminal and Git experience can be limited, and the first sessions cover setup and basics before labs build in difficulty.
Is the training tied to one AI model?
The core course teaches Claude Code. Zapper Edge is model-agnostic as a company, so we can discuss how the same governance habits apply to other tools and models your team uses.
Can the labs use our own code?
The standard labs use a prepared practice repository so nothing sensitive is needed. For a tailored engagement, labs can be adapted to your stack and policies, agreed in advance with your security team.
See it on your own documents and data
Tell us what you want to automate and we will show you how it runs inside the harness.