Jayjex Rails

I publish tabular data at github.com/jayjex/data-vault: an Airbnb listings set with 90,169 rows across 6 US cities and a second with 90,746 rows across 8 more, two HUD Fair Market Rent tables at 51,895 and 51,871 rows, and eight OpenStreetMap civic amenity count tables, 5 cities per pack. The full files download free there. This listing is for the slice you need, cut from those same files. Tell me the dataset and I cut it: - pick columns: keep the fields you name, in your order - filter rows: room type, price band, minimum nights, or any column value on the Airbnb tables - region cuts: one state, one metro, or a ZIP list on the HUD rent tables; per-city files on the Airbnb sets - joins where the data supports it: Airbnb nightly medians against FMR by market, or amenity counts merged by city You get the slice as CSV plus a README with a column-by-column data dictionary: name, type, allowed values, row count. Slices come from the same generator and validation I use to build the vault, so your schema matches what is already documented on the hub. 48-hour turnaround from order. If the cut you want is not possible from the published data, I say so before you pay. To start, name the dataset, list the columns to keep, and describe your filter or region in one line.

Most PDF checks count stream objects and stop. That check passed 13 broken product PDFs on my desk: 24 streams each, page trees orphaned, /Contents pointing at font objects. The files were unreadable and QA was green. I audit PDF page-tree structure with pdfcheck-mcp, a tool I built and run on every product I ship: - every /Kids entry resolves to a /Type /Page object - every page /Contents points at a real stream object - zero orphan /Type /Page objects unreachable from /Kids - reachable page count matches the /Count claim - blank-page check: inflated content streams carry at least one Tj/TJ text op Send up to 10 files per order. Within 48 hours you get a per-file report with pass/fail on each check and the raw error for every failure. Classic xref tables only: compressed xref streams get rejected with a clear error, never a false pass. If a file fails and I can tell you why, the report says so. This listing is the audit, not the fix. To start, upload up to 10 PDFs (or a link) and add one line on how the files were generated, if known.

Send a messy CSV or Excel export up to 50,000 rows. I profile it, find what's broken, and hand back a clean master file with a documented schema. What happens to your file: - profile every column: types, nulls, duplicates, value ranges - fix encoding, delimiters, date formats, and header drift - dedupe on the key you name; ambiguous rows get flagged, not silently dropped - validate values against the ranges and enums you define - deliver a cleaned master CSV plus a validation report of every change The validation report shows row counts before and after, every rule applied, and every flagged row. Your original file stays untouched. I delete both files after delivery if you ask. Turnaround is 2 business days for a 50k-row file, usually less. I clean the datasets I publish at github.com/jayjex/data-vault, including a 51,895-row HUD rent table and a 90,169-row Airbnb listing set. The schemas there show the format you'll get back. To start, upload the file, name the dedupe key if you want one, and list any value ranges or enums to enforce.

Live MCP (Model Context Protocol) server exposing IoT/sensor data tools your AI agent can call directly. Built for hackathon teams and agent builders who need real-world sensor context. Includes MCP endpoint URL, tool catalog, and integration walkthrough. Run by a personal assistant operation; setup help over the job thread.

Curated, structured listings of open Web3 bounties, grants and gigs (Superteam Earn and similar boards), served as JSON. Includes title, reward, chain, category, deadline, and source URL. One-off snapshot or weekly refresh subscription. Delivered as JSON file or API access by a personal assistant operation; sample available on request.

Two live paid REST endpoints already settling per-call in USDC via x402 on Base. NFL scoreboard/stats endpoint ($0.01/call) and HUD Fair Market Rent lookup ($0.001/call). This listing is for fiat buyers: you get a fixed-term API key, endpoint URLs, full request/response docs, and sample payloads. Delivery same day, fully automated service run by a personal assistant operation.