About

Why a lab?

Reading about agents is easy. Deciding when to build one, and designing it so it survives real users, real documents and real auditors, is the hard part. That only comes from practice.

What makes this different

  • You do the work. Every case study ends with you designing the solution, not with a model answer to skim.
  • Feedback against a published rubric. You see exactly what "good" looks like, and an AI grader scores your design criterion by criterion.
  • Real-world judgment, including when not to build an agent. Sometimes an API call or a plain robot is the right answer. Knowing that is a senior skill.
  • Concept-first. Orchestration, human-in-the-loop, guardrails, evals and tool design are the same ideas across UiPath, LangGraph, the Claude Agent SDK and MCP.
  • Paired with video. EachLearning by Doing video links to a lab so you can practise straight away.

Who it's for

  • RPA developers moving into agentic automation
  • Engineers preparing for agentic / AI automation interviews
  • Teams upskilling on agent design and governance

Teams & workshops

Running an RPA or automation team that's moving to agents? Workshops and 1:1 mock design interviews are in the works. Join the list below, or reach out on LinkedIn to talk about it.

Who's building this

Parth Doshi

Agentic automation architect, speaker & mentor

  • Speaker at UiPath Community Mumbai: Agentic Days & virtual sessions
  • Interviews and mentors engineers moving into AI roles
  • Runs the "Learning by Doing" YouTube channel
  • UiPath Certified: Agentic Associate & Professional
  • UiPath Certified Solution Architect
  • UiPath Certified Advanced Professional Developer
  • Claude Certified Associate: Foundations

More about the project →