AI for developer
teams
I teach developers to work with AI systematically. It’s not about writing better prompts. It’s about a method that organizes team workflow, and an entire tool ecosystem I provide: from MCP servers, through agent orchestration, to production workflows.
Running Claude Code isn’t enough to do serious AI-powered development. Real projects require a systematic approach: code review, security, documentation, and multi-tool coordination. I deliver a battle-tested working environment. The BMad method, configured tools, and standards that get the team moving from day one.
How do we work together?
I don’t run off-the-shelf trainings. Every program is tailored to the level and needs of a specific team, so we work in three stages.
Discovery
An online call with your tech leads and a short team skill survey. We agree on technologies, expectations, and technical requirements, and decide how much time to spend on live demos and how much on hands-on exercises. Based on that I tailor the program and advise whether one training day is enough or two are worth planning.
On-site training
An intensive day at your office, in a group of up to 12 people. The morning is a lecture with live demos: Claude Code from fundamentals to advanced techniques, AI working methods, and the tool ecosystem. In the afternoon participants work with Claude Code on the training repository under my guidance.
Follow-up
Two to four weeks later we meet again for a remote Q&A session. By then the team has tested the new techniques against daily work, and I help solve the real problems that came up along the way.
The training requires no access to your code, repositories, or data. We work on a training project I prepare, and the knowledge transfers one-to-one to your own projects.
Why is it worth it?
Results from day one
During the training, developers work on real projects and see how AI accelerates their daily work. They don’t wait weeks for results.
Security and control
AI doesn’t work on autopilot. I teach teams how to maintain full control over code, how to do code review with AI, and how to avoid common pitfalls.
Whole team transformation
I don’t train individuals, I transform the way the entire team works. Shared standards, shared tools, shared workflow.
A complete ecosystem, not a prompt course
I deliver a complete working environment: BMad method, configured MCP servers, agent orchestration tools, and standards that get the team going immediately.
What will your team learn?
- BMad method: structured AI project management, from planning through implementation to documentation
- Practical programming methods with Claude Code
- Ecosystem configuration: MCP servers, CLAUDE.md, hooks, and automation
- Git Worktree: parallel work with multiple AI agents simultaneously
- Effective task delegation to AI agents while maintaining control
- AI code review: how to verify and improve generated code
- Project documentation generated on the fly with AI assistance
- Security standards and governance in AI-assisted work
- Agent orchestration: coordinating multiple AI tools in one workflow
- Modernizing existing code using AI
How does the training work?
The training is delivered progressively. We start with the basics and add tools and techniques with each stage, building up to the target workflow.
AI fundamentals
We explore Claude Code. Basic commands, environment setup, first tasks delegated to AI agents. The team sees what daily AI-powered work looks like from the very first hour.
Project memory and change tracking
CLAUDE.md as the project constitution. The agent knows the rules, standards, and architecture. Todo files for tracking progress. Project context that makes AI work smarter with each task.
BMad method
A structured workflow with specialized AI agents: analyst, architect, developer, PM. From requirements analysis, through planning and implementation, to documentation. Each stage has its own agent and clear rules.
Git Worktree and parallel work
Multiple agents work simultaneously on separate branches without interfering with each other. You work on one task while the agent handles another in the background. True parallelism, not queuing.
MCP servers and extending capabilities
Playwright for browser testing, Context7 for up-to-date library docs, database and API servers. We learn to install, configure, and build custom MCP servers tailored to the team’s needs.
Code review and security
How to verify AI-generated code. Hooks, guardrails, quality standards. Prompt injection, hallucinations, common pitfalls. Full control over what gets into the repository.
Orchestration and automation
Coordinating multiple AI tools in one workflow. Automating repetitive tasks, CI/CD integration, working with multiple agents simultaneously. From a single agent to a production pipeline.
Working on real projects
We apply everything above in practice: on your team’s code or, when code access is off the table, on a prepared training repository. Real challenges, real results. Developers finish the training with a working workflow, not notes.
Follow-up and optimization
After the training, I don’t leave the team on their own. I monitor progress, answer questions, help optimize the workflow, and react to new tools and opportunities.
What’s included?
- Discovery: a call with your tech leads and a team skill survey
- One on-site training day, in a group of up to 12 people
- Training repository and materials for participants
- Remote follow-up session after two to four weeks
- Help choosing Claude licenses: the licenses stay with the team, so learning doesn’t end with the last hour of the training
Want to train your team?
Let’s talk about how your developers can start working effectively with AI.
mariusz@tarnaski.pl