Work

Four projects in more detail: the business problem, what we decided and why, and what it was worth to the client.

Case studies

An agentic pipeline that critiques its own models

A team of AI agents that builds marketing-mix models, checks its own work, and produces the ROI numbers behind $50M+ in yearly spend.

$50M+

annual client spend informed

10

interworking AI agents

AWS

delivery partner

  • Agentic AI
  • AI Architecture

An analytics platform that writes its own R

Pharma data scientists ask a question in plain English and the platform writes and runs the R. It started as an MVP and grew into a $3M engagement.

$3M

engagement grown from MVP

4

person team led

2

roadmap pivots from research

  • Generative AI
  • Platform & Workflow
  • Product Ownership

Retiring the spreadsheet a reporting process ran on

Swapped a hand-edited Excel reporting process for a layered data lake on AWS. Monthly reporting got 30% faster for 100+ business users.

30%

faster turnaround

100+

business users

  • Data Architecture
  • AWS

6,000 PDFs that nobody could safely edit

6,000+ procedure documents used to live as PDFs. Now people describe an edit in plain English, review a redline, and send it to the right regional approver.

6,000+

SOPs supported

5

regions

4

person team led

  • Generative AI
  • Platform & Workflow
  • Team Leadership

“In the land of the blind, the one-eyed man is king.”

Erasmus, Adagia

Also built

Also built

Conversational AI insights assistant

Senior commercial stakeholders waited on analyst cycles for answers. Launched an AI assistant across five business units of a Fortune 500 client, one unit at a time, proving value through user testing before expanding. It reached 95%+ insight accuracy and 90%+ monthly active use among 15+ senior stakeholders, and won the client's firmwide Innovation Award in 2025.

  • Generative AI
  • Adoption

GenAI billing classification

Billing categorization for 10,000+ employees was manual and repetitive. Defined a billing ontology with the internal finance team, then deployed a generative AI classifier against it, validated against a hand-tagged set and with human review before production. Manual effort fell 70%.

  • Generative AI
  • Automation

Metric Engine

Each business unit defined its performance metrics differently, so the numbers disagreed. Working with the engagement principal, prototyped a single source of truth for metric definitions and showcased it to multiple Fortune 500 clients. It led to a two-year engagement.

  • Data Architecture
  • Zero-to-One

Self-service data catalog

Business users at a pharma client depended on engineers to find data and change quality rules. Deployed a data catalog covering 100+ data assets, so users could discover data and set quality rules and approval workflows themselves.

  • Data Governance
  • Self-Service

Reusable reporting architecture

Every new client meant rebuilding the path from raw commercial data to reports. Built a multi-layer database system that cut onboarding-to-first-report time 30% across five deployments. It became a reusable firm asset, later integrated into a $50M commercial product.

  • Data Architecture
  • Platform

Chief, a personal career agent

A side project for MBA recruiting: a single-user agent that ingests job postings and rewrites résumé bullets for each role, drawing only on a curated file of verified facts.

  • Agentic AI
  • Side Project