ROCKI
EN简

Services

Work with me

I take small, clearly scoped jobs in three areas. The samples below are my own self-initiated work, so you can check the quality before you get in touch.

Data extraction

I build extractors that turn public web pages, documents and APIs into clean JSON, CSV or Markdown, packaged as Apify Actors you can run yourself.

AI-assisted, human-reviewed.

Self-initiated sample, not client workSample sheet for Sitemap URL Extractor: the input JSON, the result counts (116 URLs) and the first rows of the CSV and Markdown output.

Sitemap URL Extractor

Run locally on rocki.org: 1 sitemap read → 116 URLs exported to CSV and Markdown.

Open full size
Self-initiated sample, not client workSample sheet for SEO Metadata Extractor: the input with 5 rocki.org URLs, the result counts and the CSV and Markdown output.

SEO Metadata Extractor

Run locally on 5 rocki.org pages → 5 rows of page metadata (title, description, canonical, Open Graph, JSON-LD) exported to CSV and Markdown.

Open full size

Portfolio: 24 public Actors on the Apify Store, all open source.

n8n automation

I build bounded n8n workflows that take data from a trigger or webhook to a clean, deduplicated file or sheet.

AI-assisted, human-reviewed.

Self-initiated sample, not client workn8n editor canvas for Sample A: manual trigger, fetch sitemap.xml, parse XML, one item per URL, normalize fields, sort, convert to CSV and write to disk, plus a disabled Google Sheets node.

Sample A: sitemap → CSV

Pulls rocki.org's public sitemap, normalizes each URL entry (path, section, language, alternate links, dates), then sorts and writes a CSV.

Run on a local n8n instance: 1 sitemap → 116 CSV rows.

Open full size
Self-initiated sample, not client workn8n editor canvas for Sample B: Apify webhook, a run-succeeded check that ignores other events, config, get dataset items, map fields, remove duplicate final URLs, sort problems first, convert to CSV and write to disk.

Sample B: Apify webhook → deduplicated CSV

When an Apify Actor run succeeds, its webhook triggers the flow: fetch the run's dataset items, map fields, remove duplicate final URLs, sort problem pages first and write a CSV.

Tested locally with a simulated webhook and a mock dataset API: 11 items in → 8 rows out.

Open full size

AI logo to vector

I turn an AI-generated logo raster into a clean, editable SVG and PDF, with intentional nodes instead of a raw auto-trace.

AI-assisted, human-reviewed.

Self-initiated sample, not client workBefore-and-after sheet for the fictional mark Tidewren: AI-generated raster, raw auto-trace and hand-cleaned vector, with 800% edge crops showing the nodes.

Tidewren (fictional mark): auto-trace 953 nodes → hand-cleaned 210 nodes (−78%).

Open full size
Self-initiated sample, not client workBefore-and-after sheet for the fictional mark Fernmoth: AI-generated raster, raw auto-trace and hand-cleaned vector, with 800% edge crops showing the nodes.

Fernmoth (fictional mark): auto-trace 2,476 nodes → hand-cleaned 709 nodes (−71%).

Open full size
Self-initiated sample, not client workBefore-and-after sheet for the fictional mark OVQUEN: AI-generated raster, raw auto-trace and hand-cleaned vector, with 800% edge crops showing the nodes.

OVQUEN (fictional mark): auto-trace 343 nodes → hand-cleaned 79 nodes (−77%).

Open full size

Fewer nodes is not the goal by itself; the aim is clean, intentional nodes.

Contact

Send a short description of the job and the data or files involved.

XGitHub