Senior Product Manager working at the intersection of AI systems, data infrastructure, and enterprise SaaS.
I build AI products that actually get deployed — not just demoed. That means caring as much about data quality, trust architecture, and organizational adoption as about the model itself.
Most AI products don't fail because the model is wrong. They fail because the data feeding it is broken, the infrastructure can't scale, and nobody built the governance layer that enterprise buyers require before they'll let AI anywhere near their workflows.
That's the problem I solve.
My background spans:
- AI & agentic systems — designing workflows where AI autonomously reasons across steps, not just generates a single response
- Data infrastructure — ETL pipelines, data warehouses, analytics platforms; the unglamorous layer that determines whether any AI insight is trustworthy
- Enterprise platform & IAM — SSO, IdP integrations, zero-trust architectures; the access control layer that gets AI products through security review
- B2B SaaS — 8+ years shipping products at Vendasta and across the enterprise SaaS stack
| Repo | What it is |
|---|---|
| pm-research-agent | Agentic workflow that autonomously researches a company and generates a structured competitive brief. Multi-step reasoning + web search, no fixed pipeline. |
| pm-prompt-library | Curated high-performance prompts for PMs: discovery, strategy, communication, and analysis. Built from real workflows, not theory. |
| ProductTeardowns | Deep-dive analyses of AI and SaaS products — how they're built, why they're positioned the way they are, and what a competing PM should do about it. |
The best AI PMs aren't the ones who can prompt the best. They're the ones who can:
- Define the right problem — most AI initiatives fail at problem framing, not model selection
- Set up the right data infrastructure — a model is only as good as what you feed it
- Build the organizational conditions for AI to be trusted — governance, auditability, change management
- Create evaluation frameworks before shipping — "good" needs a definition before you go to prod, not after
I'm also a believer that the IAM and access control layer is the most underrated part of enterprise AI product work. I've seen more AI deals die in security review than in any product demo.
- AI evaluation frameworks — how to measure whether an AI product is actually working, not just generating output
- Agentic workflow design — orchestration patterns, failure modes, and when not to use agents
- AI security & governance — prompt injection defenses, bias auditing, auditability within IAM boundaries
Working toward my private pilot license. There's a version of product management in every pre-flight checklist — precise inputs, shared mental models, and a clear understanding of what happens when requirements are ambiguous at 3,000 feet.
"I don't just use AI to work faster — I orchestrate AI to change how work gets done."