Kevin Burke.
Hire Kevin
SELECTED WORK2026
// SELECTED WORK

Four results from one year at a media agency.

All from 2026 at Noble People, an independent agency of about 50 people, where I lead the analytics and AI team.

CASE 01NOBLE PEOPLE · 2026
// CASE 01

I rebuilt an agency's analytics function on my own, then put AI on top of it.

When I joined, the previous analytics team had left. Every client depended on reporting built on undocumented dashboards and a sprawling warehouse, so I built the system that would do the team's work.

  • One BigQuery warehouse unifying 12 ad platforms, with storage cut 96% (2.2 TB to 93 GB) and no live dashboard broken.
  • A 21-check daily monitor that catches platform breaks before clients see them.
  • Rocky (AI Agent), an AI operating system on Claude Code connected to 15+ tools.
18
active clients served by the system today
$120M
in annual media spend mapped to clients
$400K+
in video spend found missing from ad-server reporting
CASE 02NOBLE PEOPLE · 2026
// CASE 02

Client decks that took days now take hours, and every number traces to its source.

I turned deck building into one pipeline that pulls live data, scripts the analysis, builds the slides and renders every page for review. A person still reads and sends. It is now the agency's default way to build a deck.

  • Every figure comes from cached data, so any number can be traced to its source.
  • On one competitive analysis, cross-checking with Claude, ChatGPT and Gemini caught a bias in the first model, plus a metric labeled monthly that was really a four-month total.
Hours
to build a client deck, down from days
15+
data-driven client decks shipped
3
AI models cross-checking findings, an LLM evaluation step
CASE 03NOBLE PEOPLE · SUMMER 2026
// CASE 03

I took two weeks of paternity leave, and every client request still got answered.

I was the only analytics lead, with every client channel on me and nobody to cover them. In one week I built a Slack command that logged requests to BigQuery around the clock, a routine that worked the queue through written playbooks, and a daily review where I approved the drafts. Nothing sent itself.

  • A priority deliverable still landed on its due date.
  • Each gap I found became a permanent fix within a day.
0
client requests missed
15+
production releases shipped during leave
1 hr
of review a day
CASE 04NOBLE PEOPLE · 2026
// CASE 04

I wrote the AI rules leadership signed, and had a funded plan for my team within a month.

People needed to know what was allowed, what was possible and who to ask. I wrote that down and got it signed.

  • Wrote the agency's AI safety and adoption memo. Leadership approved it and funded a managed Claude plan for my team within a month.
  • Built an internal AI hub with 56 documented deliverables across 12 categories.
  • Lead a team of two analysts, both direct reports, who run agents cloned from Rocky and own 8 client accounts each.
1 month
from my safety memo to a funded Claude plan
2
analysts running agents cloned from Rocky
16
client accounts my two analysts own
CONTACTKEVIN BURKE
// HIRE KEVIN

Let's talk.

Tell me what you're building. I'll walk you through Rocky and what my analysts have done with their agents, and we can see where it fits. Email is fastest. I'm in the Philadelphia area, open to hybrid or remote.

INTERVIEW MODE · ← → to step · Esc to exit