Software engineers learn to run AI agents in production.
Free Live Labs every Thursday on Maven. Param builds one agentic system from a public repo, you build it with him, then you take it to your own code. It is an AI engineer roadmap for engineers who have shipped one LLM feature and want the next one to hold up.
Join Thursday's free labEvery lab's code is public on GitHub, one folder per lab in agentic-ai-lightning-sessions.
Watch a lab before you join one
Each past Live Lab is cut into a short mini course. Start with the first lesson here, no account needed.
Design a Software Factory for Your Coding Agents
- What a Software Factory AutomatesPlaying
- Stop Babysitting Your Coding Agent4:03
- Verification Is the Bottleneck4:47
- Turn Agent Rules Into Gates4:40
- Boundary, Context and Skills7:34
- Isolate Every Agent Run3:45
- Why Verification Is the Hard Layer6:15
- Tests First, Then Verify Outside the Agent4:31
- From Pull Request to Merge6:47
- Human Review Stays in the Loop4:50
Why do agents pass the demo and then stall?
Most engineers have built the tutorial agent. The trouble starts on the first real ticket.
Tutorials stop at the demo. A notebook has no users, no permissions and no bill. Your first production ticket asks for all three.
The agent acts differently on every run. Without traces and evals you cannot tell a fix from luck, so nobody trusts it with real work.
Nobody near you has shipped one. Your team asks how agents should work, and you are learning it alone after hours.
Each Live Lab takes one production problem, builds the fix live from a public repo and hands you the code. You leave with something running and a pattern you can apply at work.
What you get from the labs
A running system every week.
Each lab ends with code in a public repo that you clone and run the same day.
Patterns that outlast the tool.
Labs teach workflows, agents and guardrails as ideas that carry across models. Claude is used where it helps, and every pattern is shown working on its own.
Production problems first.
Security, cost, traces and evals each get a lab of their own: Secure AI Agents with Guardrails, Cut AI Agent Costs with Token Economics, Trace AI Agents in Production and Build AI Agent Evals from Production Traces.
Answers while you build.
Every lab ends with open questions, and the live-labs repo keeps GitHub Discussions open between labs.
A path when you want more.
Workshops, the Agentic AI Buildcamp and Pro recordings pick up where the free labs stop.
Start with a workflow and earn the agent.
The AI engineer roadmap, one lab at a time
Six stages, in the order production asks for them. Each stage links the lab that builds it and the guide that explains it.
Control
Build a Mission Control for AI AgentsWatch the recording
Cost
Cut AI Agent Costs with Token EconomicsThu 12 Nov, 17:00 UTC
Observe and evaluate
Trace AI Agents in ProductionThu 10 Dec, 17:00 UTC
Build AI Agent Evals from Production TracesThu 3 Dec, 17:00 UTC
Govern
Build an AI Agent Governance FrameworkThu 17 Dec, 17:00 UTC
How a Live Lab works
Sixty minutes on Maven, every Thursday at 16:00 UTC in October and 17:00 UTC from 5 November.
Save a seat on Maven
Each Live Lab is a free Lightning lesson. You sign up with your email.
Build it with Param
He frames the problem, builds the system live from the public repo and takes your questions at the end.
Take it to your own code
Clone the lab folder, run it on your repo and bring what broke to the next lab or to Discussions.
Pick the next step that fits your project
- Join Thursday's free lab
Live Labs
One system built live every Thursday.
Free on Maven with your email.
- See the workshops
Workshops
Four hours on one hard part, with your version built by the end.
$497 each on Maven.
- See the buildcamp
Agentic AI Buildcamp for Engineers & PMs
Four weeks, two production systems, 16 November to 11 December.
$997 on Maven.
- See Pro
Pro recordings
The buildcamp sessions cut into courses you watch at your own pace.
$149 a year or $399 once for lifetime.
- Plan a private lab
For teams
A private lab on your own codebase for up to 20 engineers, scoped with KaaS.team.
The next Live Labs
See all labs and recordingsOctober
Build a Mission Control for AI Agents
A board of AI employees that resolve a company's billing issues, where a refund waits for your tap and a card closes only when the payments ledger agrees.
Save a seat on MavenBuild a Multi-Agent System with MCP
A planner, an engineer and a reviewer that take one issue from plan to pull request, each with only the tools its role needs.
Save a seat on MavenNovember
Secure AI Agents with Guardrails
Agents that each sign in with their own identity, act only within the user's permissions and refuse an injected instruction.
Save a seat on MavenCut AI Agent Costs with Token Economics
One agent run priced step by step, rolled up to a cost per feature and per customer, and cut with routing, caching and a budget.
Save a seat on MavenBuild a Second Brain with Claude Code
A second brain over your notes, docs and Linear that answers with sources and serves the same memory to any agent over MCP.
Save a seat on MavenDecember
Build AI Agent Evals from Production Traces
A test set built from real agent runs, scored by code checks and an LLM judge, that fails the build when the agent gets worse.
Save a seat on MavenTrace AI Agents in Production
Every agent run traced in Langfuse, a dashboard of cost, latency and failures per agent and per customer, and an alert before one goes over budget.
Save a seat on MavenBuild an AI Agent Governance Framework
A way to decide which work goes to agents, who approves what they do, and what they changed each month.
Save a seat on Maven
Check the work before you join
- learnwithparam /agentic-ai-lightning-sessionsOne folder per labOpen on GitHub
- learnwithparam /live-labsDiscussions between labsOpen on GitHub
- learnwithparam /software-factoryThe factory from the buildcampOpen on GitHub
- learnwithparam /claude-verification-harnessThe harness from the Lightning labOpen on GitHub
“You learn a system by building it, breaking it and fixing it. So that is what every workshop does.”
- Building software since 2011.
- Architect of the pivot behind Jobbatical's €11.6M Series A.
- Led the engineering work behind ISO 27001 certification on a regulated platform.
Questions about learning AI engineering here
Is there a free agentic AI course?
Yes. Every Live Lab is a free Lightning lesson on Maven, and every past lab is a free mini course on /courses.
Who is this for?
Software engineers with a few years of shipping who have built one LLM feature and want agents that hold up in production. Engineering leads and PMs who read code join too.
Who is this not for?
People new to programming, and anyone after prompt tips more than systems. If you cannot run a Python or TypeScript repo yet, start there first.
Which tools and models do the labs use?
Claude and Claude Code appear where they fit. Each pattern is built so you can swap the model, and the repos show where the swap happens.
What if I miss a live lab?
The recording stays on Maven after the session, and the mini course on /courses follows.
Who owns the code I write?
You do. Everything you build in a lab, a workshop or the buildcamp is yours to keep and use at work.
What does it cost?
Live Labs, recordings and the newsletter are free. Workshops are $497 each and the buildcamp is $997, paid once on Maven. Pro is $149 a year or $399 once for lifetime.
Can my company pay?
Yes. The team kit has a reimbursement email to send your manager and an invoice in your company's name. See /for-teams.
Do you train whole teams?
Yes. A private lab runs on your own codebase for up to 20 engineers, scoped with KaaS.team.
What is settled before you join
A Live Lab costs nothing and needs only your email.
The recording stays on Maven after the session.
The code is yours to keep, in a public repo you can fork.
Paid programmes are one payment on Maven, priced the same worldwide.
Build one system with us this Thursday
Join Thursday's free lab
Rather read first?
The Top 1% Engineers newsletter sends one production lesson per issue.
We email you each new issue. You can leave at any time. Privacy