Military veteran. 11+ years problem solving with teams in high-risk environments; now applying that same problem solving to messy operational data.

I sit with the real workflow, pull signal out of unstructured text and production systems, and ship tools people actually use.

Aim: Embedded delivery — sit with the workflow, pull signal out of messy operational systems, put the answer where the team already works.

• Problem framing in the field — find the pain, name the decision, ignore theater
• Text + ops pipelines — Python, SQL & NLP on noisy operational text; same problem solving on whatever system is generating the mess
• Ship and measure — cut cycle time, put a number on the result, kill work that does not pay off

• Spike RCA pipeline — Replaced a 50-minute manual review of speech-analytics alerts with a Python pipeline (intelligent sampling → full transcripts → structured RCA). Turnaround 12–15 minutes (~70% faster), deeper sample, same-day answer for ops leadership. Public repo is a synthetic reconstruction of the method, not employer data.
• Call-driver text mining — Beyond keyword volume: scored client phrases in mid-February 1099 transcripts (Fightin’ Words + c-TF-IDF) so leadership could see why contacts spiked. Same-season week-vs-week contrast; packet for FAQ + IVR copy. Public repo is a synthetic reconstruction of the method, not employer data.
• Consumer Light rate card — Tested whether a $9.99 consumer shipping rate covered a 40 lb DIM UPS Ground parcel. Zone join from origin 84098 showed the book sat in zones 5–8; the move to $14.99 cut the per-box hole by about $5 and did not close those zones. Public repo is zone aggregates and discount bounds, not order-level data and not an invoice.