Work

AI that made it to production, and the demos I build now

The first part of this page covers AI tools I led while Director of Data Science and AI at a Fortune 500 equipment rental company. They are examples from my career as an employee, not client engagements. The second part shows demos I have built since.

Part one

Production AI inside a large enterprise

Each followed the same path: my team proved the value with a proof of concept and MVP, an engineering partner or IT took it to production, and we measured use after launch.

Troubleshooting assistant for field technicians

In productionAWS Bedrock

Problem
More than 5,000 field technicians lacked fast access to troubleshooting and repair knowledge, which slowed repair cycles.
What was built
My team built the proof of concept and MVP on AWS Bedrock. An AWS services partner then delivered and scaled the production version.
My role
I prioritized the use case, led the proof of concept and MVP, set the evaluation and human-in-the-loop standards, handed the build to the partner, and tracked adoption after launch.
Result
About 70% adoption within six months across technicians and managers.

Agentic assistant for corporate and field teams

In productionAWS Bedrock and AgentCore

Problem
Managers needed a single place to ask policy questions, look up equipment specs, find people and roles, and see which branch serves a given ZIP code.
What was built
An agentic workflow with retrieval over policy documents plus structured lookups, delivered to production by my team together with IT.
My role
I led delivery from use-case definition to production and set the architecture pattern and the governance and human-in-the-loop standards it runs under.
Result
About 2,000 monthly active users. Several manual lookup processes replaced by one assistant.

Equipment recommendation agent for customers

Live for customersAWS Bedrock

Problem
Customers needed help finding and comparing the right equipment for their project without depending on a sales rep.
What was built
My team built the proof of concept and MVP. An AWS services partner delivered the production version, rolled out to employees first for testing, then to customers.
My role
I conceived the use case, led the proof of concept and MVP, and managed the handoff to the partner and the phased rollout.
Result
Guided equipment selection, now live for customers online.

Plain-English questions on existing KPI dashboards

In productionSnowflake Cortex

Problem
KPI dashboards were already built and automated, but frontline users without technical training needed an easier way to ask questions and drill into details beyond the standard views.
What was built
A semantic layer on top of the existing dashboards that answers questions in plain English. Because the underlying data was already automated, no data cleanup was needed first.
My role
My team partnered with the BI team and trained them to build and run it.
Result
Frontline, non-technical users can drill into KPI data themselves in plain English.

Before the AI work

The analytics foundation underneath

Before leading AI, I spent a decade building the analytics function at the same company: KPI definitions, governance, shared data definitions across functions, and a scalable data foundation of pipelines, curated summary tables and automated Tableau and Snowflake dashboards that let business teams measure performance themselves. When AI arrived, that foundation made it possible to move quickly.

Repair triage and technician performance analytics got needed units out of repair and back on rent faster. That work later shaped the field-service AI use cases above. It is also why I start AI projects with the data and the workflow, not the model.

Part two

Demo

Working prototypes I built to show an idea in practice. Where a demo needs a login, I will walk you through it on a call rather than share credentials.

ABC Inc. Insights

A sample HR KPI dashboard that pushes what changed, why it matters, and the next best action, instead of waiting for someone to open a report.

ABC Inc. is a fictional company. All data is made up.

Open the demo (opens ABC Inc. Insights)

Have a use case in mind?

Bring it to a 30-minute call. I will tell you how I would test it, and what would need to be true for it to reach production.

Email me to set up a 30-minute call