Conner
Toryfter.
Data scientist and applied mathematician on Eaton's Workforce Insights team, fresh out of UW–Madison. I'm drawn to the practical edge of this field, where careful math earns its keep against the numbers a business actually watches, like the production lead-scoring model I shipped that cut manual review labor by 80%.
| Production ML | 70 |
| Statistical rigor | 65 |
| Data visualization | 70 |
| Optimization | 60 |
| Baseball IQ | 80 |
| OFP | 70 · under contract |
On the
record.
I'm past the notebook stage. The work I find most rewarding is the full loop: framing the problem with stakeholders, training and tuning the model, getting it into production, then watching the metric move. I've done it once already at an enterprise partner, cutting manual review labor by 80%. Now I'm running that loop on workforce data at Eaton.
Academic depth backs that experience. UW–Madison's Data Science degree, with a certificate in Mathematics, built fluency across the stack: deep learning in PyTorch, reinforcement learning on Gymnasium, CRF sequence labeling, QCQP optimization in Julia, and the full statistical inference toolkit. I read papers in my downtime and publish original research on the side.
Where I am now: Associate Analyst, Workforce Insights at Eaton, on the Workforce Analytics team, working remote from Milwaukee. The job is labor market analysis across Eaton's global footprint: external benchmarking, internal segmentation, workforce planning, and the dashboards HR and leadership use to make the call. Same habits as everything else on this page: start from the decision, check the data before trusting it, and show the work.
"He gets on base." Peter Brand, Moneyball (2011)
The
evidence.
Eight projects with the highest signal: production deployments, original research, and advanced coursework with measurable results I can walk through number by number. Every number below is real. The full catalog of 31 lives in the chart that follows.
Every project,
plotted.
Thirty-one projects spanning foundational coursework through production AI, plotted by year and domain. Scroll to read the arc, or take the wheel with the filters. Hover any point for the title, click to open the file.
Thirty-one projects, four seasons.
Every dot is a real, finished piece of work — coursework, research, and production systems. Read it like a development arc: low-A ball on the left, the majors on the right.
Foundations first.
Intro Python on real datasets — 50K Airbnb listings, 100K IMDb rows, NOAA storm records. Unglamorous, necessary, fast.
The deep end of the curriculum.
PyTorch CNNs, reinforcement learning, CRF sequence labeling, QCQP optimization in Julia, permutation tests built from scratch. Method over demo.
Production, with receipts.
Real company, real stakes: a deployed lead-scoring model that cut manual review labor 80%, and an inspection app that replaced a paid SaaS contract company-wide.
The big leagues.
A full-time seat on Eaton's Workforce Insights team, plus live systems still running on the side: a 3D pitch-tunneling tool covering all 596 qualified MLB pitchers, xRAPA (an original MVP metric), a production movie recommender, a quant trading bot sizing real money. Click any of them.
The
stack.
The
institution.
Ask the
Bureau.
Don't take the file's word for it — interrogate it. This line is answered by a Retrieval-Augmented Generation system I built: your question is embedded, matched against a vector index of my résumé and all 30 projects, and answered by Llama 3.3 with citations you can open. The demo is the skill.