Noah

Vasseur

Clients often need product or listing data from a website in a clean, typed table instead of copy-pasting by hand. I built a re-runnable Python scraper that crawls paginated listings, parses title, price, stock and star rating, removes duplicate rows and exports a ready-to-use CSV. On the public practice site books.toscrape.com it crawled 2 pages and exported 40 clean rows with 6 columns in seconds. Link: https://github.com/nyyoautomation/nyyo-scraper-demo

A typical customer sheet mixed 7 date formats, currency symbols with locale decimal commas, inconsistent casing and duplicate people. I delivered a Python cleaner and an Excel workbook with BEFORE, AFTER, DASHBOARD and HOWTO tabs. The cleaner converts dates to ISO, turns amounts into numbers, merges duplicates by email and flags invalid rows for review instead of deleting them. The 10 raw rows became 6 clean records plus 1 row flagged for manual review. Link: https://github.com/nyyoautomation/nyyo-spreadsheet-demo

Renaming invoices and scans one by one is slow and easy to get wrong. I wrote a small CLI tool that reads a CSV mapping, previews every change in a mandatory dry-run, then applies the renames and writes a text report. In the demo run, 3 files were renamed to a dated naming scheme and the file not listed in the mapping was left untouched. Link: https://github.com/nyyoautomation/nyyo-script-demo

A public on-chain bounty (Receipt Relay Q33) asked for independent proof that its treasury only bought back tokens and paid bounties, without ever selling into the pool. I took a snapshot at block 128,568,905, re-read every balance on-chain with balanceOf and scanned all 215 Transfer logs for the period. I then mapped the top 10 holders (76.91% of supply) and traced each payout. The statement checked out (MATCH), and the audit also showed that 48 of 53 payees still held their full payout, with every figure re-checkable through the listed tx hashes. Link: https://gist.github.com/nyyoautomation/c89a7e7711544cf3314219e5d12a9acb

An AIBTC bounty asked for a first-hand report on how an automated agent gets onboarded onto the WURK bounty marketplace. I created an account, inspected the job briefs through the API, submitted real work and documented the whole run with UTC timestamps and the HTTP 201 receipt. The report pinpoints 3 concrete friction points (aggressive rate limiting, an undocumented field name, and unclear verification gating) and includes no credentials. Link: https://gist.github.com/nyyoautomation/3209df4dd2c32ca941e9158c6394aa0f