Workshop Lead & Curriculum Designer · Microsoft
AI Azure Engineer Workshop Series
Designed and led a multi-session workshop series with Microsoft turning software engineers into AI engineers — teaching them to connect Azure OCR to LLMs and build document pipelines that read W-2s, pay stubs, and PDFs.
In partnership with
- Microsoft
- Microsoft Azure
- Azure OCR / Document Intelligence
- LLM Integration
- Prompt Engineering
AI Azure Engineer Workshop Series
Every software engineer has watched the job titles shift under their feet: the postings say “AI engineer” now, and nobody handed out a map for getting from here to there. Through 2025, Marlon partnered with Microsoft on a multi-session workshop series built exactly for that gap — software engineers who already ship code, ready to ship AI.
Documents are the on-ramp
The series is anchored in the least glamorous, most universal problem in business software: documents. W-2s, pay stubs, PDFs — the paper that runs payroll, lending, taxes, and HR. It’s the perfect training ground for a new AI engineer because it’s real: messy scans, inconsistent layouts, and stakes that punish sloppy output. If you can build a system that reads a stack of pay stubs correctly, you’ve learned skills a demo chatbot will never teach you.
Read, then reason
The core pattern the series teaches is a clean division of labor:
- Azure OCR reads. Document Intelligence extracts the raw text and structure from whatever arrives — a crisp PDF or a phone photo of a W-2.
- The LLM reasons. The extracted output flows into a language model that interprets, validates, and answers questions about it: What’s this person’s year-to-date income? Does this stub match that W-2?
Engineers build the full pipeline hands-on, session after session — connection by connection, not slide by slide.
The model is the easy part
The hard part of the series was convincing experienced engineers of an uncomfortable truth: calling the LLM is the easiest step in the pipeline. The real AI engineering is everything around it — handling OCR output that’s almost right, deciding what the model should see and in what shape, and knowing when the system should say “I’m not sure” instead of guessing about someone’s paycheck. The sessions spend their depth there, trading tool-tour breadth for one pipeline engineers actually understand end to end.
Why it matters
The AI engineer shortage isn’t a talent problem — it’s a bridge problem. The engineers are already here; they just need a route from the systems they build today to the ones companies need tomorrow. This series is that route, and the pattern it teaches — read with OCR, reason with an LLM — is the same one running inside half the AI products shipping right now.
The flywheel
This build became a lesson.
Every system Marlon ships turns into teachable patterns — in keynotes, workshops, and executive sessions. Want the lessons from this one, live?