Sparround

Portfolio: what to show and how to talk about it

For n8n roles it is not certificates but work you can show that decides. Answering "what have you built?" with a concrete example beats any line on a CV.

A good portfolio is three or four workflows, and they must not repeat each other. The goal is not quantity but demonstrating distinct skills:

1. A classic integration — trigger, API, data transformation, notification 2. A reliability piece — error handling, idempotency, a cursor 3. An AI scenario — a chain or an agent, with output validation 4. Modular design — split into reusable sub-workflows

Together they say: "I can build integrations, I think about production, I use AI properly, and I have a sense of architecture."

PieceWhat it demonstratesAn example idea
Classic integrationThe core skills: triggers, HTTP, expressions, notificationA daily report: API → sheet → channel
Reliability pieceProduction thinkingA cursor-based sync: cursor, upsert, error workflow, partial failure
AI scenarioThoughtful use of AIEmail classification: a chain, output validation and a human queue
Modular designA sense of architectureA shared "enrich" sub-workflow used by three workflows

How to present a piece. Sharing your screen and walking through nodes one by one is a weak presentation. Prepare a four-sentence account for each workflow instead:

1. The problem — what was done by hand, and how long it took 2. The solution — the trigger, the main steps, the outcome 3. The difficulty — what went wrong and how you solved it 4. The result — with a number where possible

The third part matters most. "It all went smoothly" signals either inexperience or lack of preparation. Something always breaks in real workflows — and that is exactly where the interviewer sees how you think.

What to do without real data. With no work experience, automate your own problem: a bot that watches an RSS feed and sends a digest, personal expenses written into a sheet, job postings collected automatically. What matters is that it genuinely runs and solves a genuine problem — a synthetic "tutorial example" is spotted instantly.

Clean a workflow before sharing it. An exported JSON does not carry credential values, but it does carry keys typed into node parameters as plain text. Real customer names, emails and internal URLs can also linger in sticky notes, node names and pinned data. Always prepare a portfolio piece with synthetic data.

📚 Sources and documentation