I help People teams move from scattered AI experiments to adoption that is useful, governed, and measured. Most of that work is deciding where AI belongs, where it doesn't, and what evidence a system needs before it earns more autonomy.

I bring HR operating experience from KPMG, MongoDB, VMware, Thomson Reuters, and SAIC, and I build working prototypes myself with Claude Code.

Portfolio · LinkedIn

• Start with the decision, not the tool. Pick use cases on value, risk, and readiness.
• Govern the design, not just the launch. Legal, Privacy, and employee representatives see it early.
• Earn autonomy with evidence. A system gets more room only after it passes tests it wasn't built on.
• Humans make consequential employment decisions. AI informs hiring, pay, and performance. It does not decide them.
• Measure changed work, not launched tools. Training and measurement are in the plan from day one.

1. Work redesign: the People Partner role One HR role broken into tasks and sorted into rules, AI assistance, and human judgment. The finding: no task yet meets the bar for a model acting on its own.

2. Resolve An HR case workflow with policy citations, a human approval gate, and an audit trail. It is rules-based, with no language model. It passes all 60 regression cases and 0 of 16 held-out cases, so production use is withheld. I publish that gap on purpose: it shows what an HR agent has to prove before anyone trusts it with real employees. Try the demo (it runs in your browser) or read how it works.

3. HR AI Transformation Playbook The operating model behind both: 37 use cases with a prioritization matrix, governance templates for the EU AI Act and US state laws, evaluation tooling, a literacy curriculum, and a live ROI calculator.

All data in these projects is synthetic. They are prototypes and reference designs, not legal advice or production HR systems.