A US hotel-intelligence company. 2024 to 2026.
A US hotel-intelligence company needed market data collected every night at a volume its team could not watch by hand. Python collectors with proxy rotation write into the client's MariaDB through stored procedures, and the engineers we placed watch them.
Screens from this system are being added.
A US hotel-intelligence company ran its competitor-rate collectors on self-hosted servers and checked its data by hand. It needed Python scraping capacity that stays reachable against sites that block automated traffic, engineers it could trust with that work, and, from early 2026, an outbound engine and a marketing site it could own.
API-first Python collectors with residential and mobile proxy rotation, retry and anti-blocking strategy, writing results into the client's MariaDB through stored procedures
Scheduled collection and monitoring for a pipeline that has run every night since the 2025 contract, and the operational work of keeping it healthy
Engineers placed on the collection work after a documented screening task, a 24-hour reverse-engineering exercise with no Selenium allowed, and a paid trial, with a backup pool behind them
A same-day property review, May 2026, briefed and delivered inside one business day
Browser research agents that worked through the client's property list, resolving booking-platform identifiers, rebrand status and opening dates
8,869 contacts, cleaned address by address and filtered from tens of millions of records down to the US companies and people worth writing to, with a bulk classification pass over 9.44 million rows across nine files feeding it
A marketing-site rebuild as a Next.js front end over headless WordPress, deployed on Vercel in April 2026
A technical discovery and security audit of the client's legacy Laravel portal, June to August 2026, with a takeover tracker, a controller security review and a deployment-validation runbook gated on a tested restore
An NDA, a screening phase and a paid-trial phase, with recorded interviews
The client has a collection pipeline that stays up against hostile sites at a fraction of what the third-party feeds it replaced were costing, a bench of engineers it did not have to find itself, and a web estate it can take over on its own terms.
Staff augmentation scoped for the client's collection work: the screening process written and the first two candidates presented.
Trial kick-offs for the placed engineers.
Fixed-price Python scraping contract begins, structured as fortnightly milestones, and converts to a retainer in September.
Prospect list built for the client's outbound, with the first outreach sent on 20 March.
Marketing site built as a Next.js front end over headless WordPress and deployed on Vercel, the phased implementation opened as a single pull request on 18 April.
A property review briefed and delivered within the same business day.
Technical discovery and security audit of the legacy portal delivered on 14 June; the deployment-validation and takeover runbook followed on 12 July, with the backend-support hiring pipeline running alongside into August.
Our own product
Our own list-cleaning engine: DNS and multi-stage SMTP checks on every address, and one of four verdicts with the reason behind it instead of a single opaque score. 34 reason codes, so every verdict points at the check that produced it.
Our own product, run on our own pipeline first
Lists cleaned before a single send, mailboxes warmed on their own domains, sequences written per persona, and a deliverability read every week. Run on our own pipeline first, then for a partner under their brand. About 225K unique recipients a month, four touches each, July to October 2025.
Our own engineering
A scheduled Python pipeline that reads Costco product pages, stores them in MongoDB and creates and updates the matching Shopify products, prices and inventory by SKU.
Elsewhere on the site
Where to next
Say so in a few sentences. The reply comes from the person who built this one.