Python that survives QA/QC — the stack, the ship list, and the iron rules.
Your existing edge — port it in week 1: openpyxl chart-generation (40-intersection volume charts), S4A crash-cleaning scripts, Overpass/OSM query patterns, ArcGIS REST pulls, KMZ parsing. Each becomes a code-script note with run instructions + a verified-against-manual date.
Stack for this job: pandas/numpy · openpyxl (+ xlwings when Excel-live needed) · matplotlib (agency-style figure template!) · geopandas/shapely/folium · requests/BeautifulSoup (respect robots/ToS; public data only) · pdfplumber/camelot for PDF tables · python-docx for report shells · pathlib/os for file ops.
Ship list: Starter Automation Projects — no longer all aspirational: the first tool (the FTO AADT puller) is shipped and in the repo; see that note’s Shipped section.
Iron rules: raw data immutable (_RAW never edited) · scripts print their assumptions · every automated output hand-verified once against manual method before trusted · code that touches deliverables gets a second pair of eyes like any calc · client data never leaves approved environments (AI Usage Rules applies to code assistants too).