Used Car Dealer, Southeast US
Overnight triage of a full auction run list. Every lot in the sale is pulled, the dealer's own buy box is applied, each surviving VIN is enriched with title status, Carfax, AutoCheck, MMR and book values, a note is written per car, failing cars are dropped with their reason logged, and the rest are pushed into the auction's own watch list ranked by margin — all before the desk opens.
A typical sale ran four to five thousand lots and the buyer had under two hours before the first lane. He worked the list top-down until the sale started, which meant roughly the last eighty percent of it was never looked at. The profitable cars were not politely waiting in the first four hundred rows. On top of that, every car he did reach needed three more tabs — Carfax, AutoCheck, an MMR lookup — and a note typed by hand, so the deeper he went the slower he got.
Built a scheduled overnight pipeline against the dealer's own auction accounts. Stage one reads the complete run list rather than a sample. Stage two applies the buy box as configuration, not code — year bands, mileage ceilings by make, a condition-grade floor, clean-title-only, no announced frame — which on a typical sale removed about two thirds of it. Stage three enriches every surviving VIN with the lookups the buyer was doing by hand, at a request rate that does not hammer the accounts. Stage four writes a note per car in the desk's existing wording, ending with the maximum bid that holds their margin after recon and fees. Stage five drops cars that fail the margin or history rules, logging the reason for each so the calls can be audited rather than trusted. Stage six pushes the survivors into the auction platform's own watch list and emails a ranked copy.
A 5,000-lot sale now arrives as roughly 800 cars, each annotated with a max bid, in the watch list by 6 AM. The buyer starts the day bidding instead of reading. Cars from the back half of the run list — previously invisible — now account for a meaningful share of what gets bought.
Python, Selenium, Playwright, Pandas, Cookie management, Cron scheduling, SMTP