Contact
Open to senior data science and applied AI roles.
Remote, or on site around Toledo, Ohio. If you are hiring, the quickest way to judge the work is the evidence rather than the writeup: every case study names the baseline it was measured against and which way the result came out. Seven of the twenty comparisons went against the build, and those are on the site too.
If you are screening me
Three minutes, and what to read in them.
Reviewers arrive with a stack of tabs and very little time. Rather than ask you to guess where the substance is, here is the order I would read this site in if I were assessing someone else.
- Judging the evidence. The work index opens with a count of every comparison measured against a baseline and which way each came out. It is the fastest read on whether I report results honestly.
- Judging statistical method. Fair Lending Lab runs preregistered hypotheses over 41,287 real HMDA records, with effect sizes, bootstrap intervals, multiplicity correction, and an explicit statement of what the analysis cannot claim.
- Judging engineering. Dispatch Optimizer is an OR-Tools CP-SAT model with a live demo, and it reports that at the time budget it actually ships the solver lands behind a greedy heuristic.
- Judging judgment. Water System Risk Index does not beat the one-feature baseline it was benchmarked against. It is published with that result on the page rather than quietly retired.
Every demo linked from this site is live and every repository is public, so nothing here needs to be taken on trust.