Brett
Bartrum

Brett Bartrum smiling in a bike helmet with a patterned visor

Platform and AI infrastructure engineer, St. Louis

I love hard problems, and I solve them so they stay solved. For 20 years I've built the systems other people depend on. Now I build ones that also get better on their own.

Open to Staff and Principal engineering rolesAt KubeCon in Salt Lake City, November 9–12

Save my contact brett@bartrum.family Résumé (PDF) LinkedIn

Twenty years of uptime

Each bar is a year. Tap one to see what I was working on and what happened that year.

Places I've been

A lot of these came from work. MiTek, TKC Global and Dell sent me across the country, and sometimes into Canada, to help organizations with their systems. Every dot is a town or city in my photo library; pick a year on the timeline above and that year's trips light up.

37
states, plus D.C.
200
towns and cities, from Kanorado, Kansas to Honolulu
5
places beyond the 50 states: Niagara Falls, Ontario, Chichén Itzá, Mexico, Costa Rica, Puerto Rico, U.S. Virgin Islands
10+
years on the road at 100+ hotel nights a year, enough for Marriott Bonvoy Lifetime Platinum Elite
Places I've traveled A map of the United States with 37 states and D.C. shaded and a dot for each of 200 towns and cities I've photographed, from Seattle to Hilton Head Island and from Anchorage to Honolulu. Home is St. Charles, Missouri. Chicago, Illinois Indianapolis, Indiana Carmel, Indiana Noblesville, Indiana Corydon, Indiana Richmond, Indiana Terre Haute, Indiana Greenfield, Indiana Marshall, Indiana Bloomingdale, Indiana West Harrison, Indiana Evansville, Indiana Elkhart, Indiana Lawrence, Kansas Colby, Kansas Hays, Kansas Goodland, Kansas Kanorado, Kansas Salina, Kansas Hutchinson, Kansas Emporia, Kansas Gardner, Kansas Bonner Springs, Kansas Abilene, Kansas Detroit, Michigan Dearborn, Michigan Rochester, Michigan Auburn Hills, Michigan Romulus, Michigan Cincinnati, Ohio Columbus, Ohio Dayton, Ohio Sandusky, Ohio Mason, Ohio Englewood, Ohio Cleves, Ohio Woodbourne-Hyde Park, Ohio Riverside, Ohio Blue Ash, Ohio Harrison, Ohio Huber Heights, Ohio Hawn State Park, Missouri Prestonsburg, Kentucky Salyersville, Kentucky Martin, Kentucky Coal Run Village, Kentucky Morehead, Kentucky Ashland, Kentucky Lexington, Kentucky Georgetown, Kentucky Louisville, Kentucky Watterson Park, Kentucky Grand Rivers, Kentucky Kuttawa, Kentucky Reidland, Kentucky Henderson, Kentucky Erlanger, Kentucky Independence, Kentucky Bentonville, Arkansas Devil's Den State Park, Arkansas Orlando, Florida Merritt Island, Florida Cape Canaveral, Florida St. Augustine, Florida Atlanta, Georgia Brookhaven, Georgia College Park, Georgia Dunwoody, Georgia Chamblee, Georgia Lookout Mountain, Georgia New Orleans, Louisiana Slidell, Louisiana Gretna, Louisiana Chalmette, Louisiana Arabi, Louisiana Hattiesburg, Mississippi Grenada, Mississippi Pearl, Mississippi Courtland, Mississippi McComb, Mississippi Seminary, Mississippi Oklahoma City, Oklahoma Tulsa, Oklahoma Broken Arrow, Oklahoma Ardmore, Oklahoma Catoosa, Oklahoma Davis, Oklahoma Weatherford, Oklahoma Muskogee, Oklahoma Moffett, Oklahoma Memphis, Tennessee Nashville, Tennessee Austin, Texas South Padre Island, Texas New Braunfels, Texas Dallas, Texas Charleston, West Virginia Morgantown, West Virginia Fayetteville, West Virginia Ansted, West Virginia Triadelphia, West Virginia Rupert, West Virginia Teays Valley, West Virginia Hilton Head Island, South Carolina Asheville, North Carolina Phoenix, Arizona Scottsdale, Arizona Glendale, Arizona Gilbert, Arizona Litchfield Park, Arizona Waddell, Arizona Casa Grande, Arizona Tucson, Arizona Green Valley, Arizona Sahuarita, Arizona Picacho, Arizona Willcox, Arizona Flagstaff, Arizona Kingman, Arizona Payson, Arizona Wickenburg, Arizona Holbrook, Arizona Heber-Overgaard, Arizona Peach Springs, Arizona Cordes Lakes, Arizona Black Canyon City, Arizona Kayenta, Arizona Anaheim, California Newport Beach, California San Francisco, California San Carlos, California Oakland, California Denver, Colorado Aurora, Colorado Greenwood Village, Colorado Boise, Idaho Horseshoe Bend, Idaho Las Vegas, Nevada Paradise, Nevada Winchester, Nevada Boulder City, Nevada Albuquerque, New Mexico Las Cruces, New Mexico Santa Rosa, New Mexico Tucumcari, New Mexico Gallup, New Mexico Moriarty, New Mexico Green River, Utah White Mesa, Utah Richfield, Utah Beaver, Utah Seattle, Washington Bellevue, Washington Redmond, Washington Snoqualmie, Washington North Bend, Washington SeaTac, Washington Burien, Washington Clyde Hill, Washington Yarrow Point, Washington Mountlake Terrace, Washington Cheyenne, Wyoming Lusk, Wyoming Washington, D.C. Baltimore, Maryland Gaithersburg, Maryland North Bethesda, Maryland Potomac, Maryland Derwood, Maryland Hagerstown, Maryland La Vale, Maryland Garrison, Maryland Millersville, Maryland Pikesville, Maryland Boston, Massachusetts New York, New York Niagara Falls, New York Philadelphia, Pennsylvania Hershey, Pennsylvania Lancaster, Pennsylvania Erie, Pennsylvania Camp Hill, Pennsylvania Somerset, Pennsylvania Middletown, Pennsylvania Intercourse, Pennsylvania Millersville, Pennsylvania East Petersburg, Pennsylvania Anchorage, Alaska Homer, Alaska Fishhook, Alaska Honolulu, Hawaii Kailua, Hawaii Kaneohe, Hawaii Pupukea, Hawaii Kahuku, Hawaii Wahiawa, Hawaii Mililani, Hawaii Waipio Acres, Hawaii Ahuimanu, Hawaii Maunawili, Hawaii Home

Shepherd, the platform I built alone

When I moved back to Missouri from Denver, no local shop would work on my e-bike. So in 2020 I opened one: Juiced Wheels, a full-service e-bike shop in St. Charles, now rated 4.8 stars across 185+ Google reviews. It grew fast. I’m neurodivergent, some say twice-exceptional, and I do my best work on systems, not the sales floor, so a team runs the shop day to day and the community keeps it busy. To run it I built Shepherd, a production AI platform for repair shops, as its architect and only engineer. It runs on Cloudflare Workers, Durable Objects and 34 durable workflows. See it live.

3,575
repair manuals behind an AI assistant, with 2,175 articles and a diagnostic knowledge graph
126,714
parts from 82 sources, searchable in about 200 ms, with Amazon price history
4,594
bike shops on a map I built with Google Maps and deck.gl
1,918
merged pull requests this year, with 1,130 test files

It also has one inbox for texts, email, Messenger and calls, with video huddles and an AI voice agent, and a daily industry podcast that writes, voices and publishes itself. It improves on its own: daily jobs find stale or missing articles and propose fixes, solved customer conversations become new articles, and every AI change has to pass measured quality checks.

I also built the tools my AI agents use for email and phones, as MCP servers. The mail server searches a 269,000-message archive across Microsoft 365 and Gmail, with hybrid vector search and 7.5 GB of attachments kept in R2, and gives each staff member access only to the mailboxes they're granted.

Problems I've solved

126,714 parts

Sorted a catalog that had no shared categories

Four distributors and dozens of brand stores each named their categories differently, or not at all. I wrote a 139-category taxonomy and had Jev, a classification model, place every part, using the vendor’s own category as evidence. 94.6% were placed automatically with high confidence, and the rest go to a review queue instead of the wrong shelf. In a hand-checked sample, 0.4% were misplaced. Jev also reads each listing’s specs into filters, and answers “not stated” rather than guessing.

$140K back

Caught Microsoft overbilling us

I built a cost platform for Azure that has tracked more than $15 million of spend since 2021, across 4 tenants and hundreds of subscriptions. It rolls spend up from business unit to subscription, resource group and resource, with a timeline, Power BI forecasts and anomaly alerts, and every resource links straight into Azure. It beat every FinOps tool we tested, and it caught Microsoft overcharging us: we got more than $140,000 back. When Claude arrived I did the same for AI. I help run our Claude Enterprise subscription, now more than 2,000 active users, with spend tracked by user, by org and by every facet Anthropic’s API exposes, piped into Datadog for monitoring.

43% → 90%

Made repair guides recommend parts that fit the job

Every repair article listed parts found by keyword search, and when I checked them, only about 43% were right: chargers on a battery-repair guide, road wheels on e-bike wheel guides. Now Jev judges each candidate against what the article says. About 90% are right, measured on hand-labeled articles and confirmed by a second, blind reviewer.

21 s → 0.3 s

Fixed a login that timed out

I inherited an undocumented API gateway: 153 APIs across about 30 environments. I rebuilt it as code and found the routing that was blocking a new environment. Login went from a 21-second timeout to 0.3 seconds.

0 logins

Took the people out of deployments

I put every deployment and release on GitOps, automated with PowerShell and Octopus Deploy, and automated new Windows servers from bare metal to app stack to monitoring. Nobody had to log in to a server to ship code again.

150 → 10

Scripted a helpdesk queue down, early on

At my first IT job, a school district’s open helpdesk tickets sat around 150. I wrote scripts for the problems that kept coming back, and the queue dropped to about 10. It’s the same habit twenty years later: fix the cause, then automate it so it stays fixed.

What I bring to a team

Ownership from end to end. At my day job I own the delivery and monitoring pipelines for 30+ environments, from architecture through on-call and production support, measured against SLOs. I find the real cause, fix it properly, prove the fix holds, and write it down for the next engineer.

Engineers who can run it without me. I mentor the people I work with every day, and I hand over systems with runbooks so the next engineer can run them on their own.

AI that people can trust. At my day job I shipped an AI-assisted change-approval system for all of Applied Systems that passed InfoSec and compliance sign-off: rules make the decision, the AI advises, and every control is tested.

Automation that removes the manual work. I've taken deployments from people logging in to servers to GitOps, automated from bare metal through the app stack to monitoring.

A different way of seeing problems. I’m neurodivergent (some say 2E, twice-exceptional). I notice the pattern other people step past, I go deep until the real cause turns up, and I build systems that don’t need babysitting.

A business owner's judgment. I run a shop with payroll and customers, so I weigh cost, risk and the people who use what I build.

Speaking and recognition

Warren Buffett’s visit with our IT team at MiTek.
TKC Global’s Overachiever Award, 2011.
Microsoft’s Redmond campus, for the Windows Server 2019 train-the-trainer.
My Dell badge, 2013.
2026
OPO Startups Tech Talk at Confluence Tech Hub: “Build Smarter Search with AI on Cloudflare’s Edge,” on hybrid search and AI agents, with a live demo of Shepherd (slides)
2025
OPO Startups Demo Day: one of four finalists with Shop DAWG, 3rd Place and People's Choice
2023
Ethical Society Mid Rivers: a talk on today's AI, how it works and its ethical implications (recording)
2018
Presented Windows Server 2019 to Dell EMC's consulting community, after Microsoft's train-the-trainer in Redmond
2015
Dell Global Support and Deployment Bronze Award, for the strategic Monsanto OS Modernization project
2012
Taught three SCCM 2012 classes to enterprise IT teams before the product shipped
2011
TKC Global Professional Services Overachiever Award, for propelling our team into new competencies, providing outstanding service, and going above and beyond
2010–12
Customer liaison for TKC Global at a major Microsoft conference
2008–10
Congratulated by Warren Buffett on the IT team’s work integrating the companies MiTek acquired
2007–08
MOREnet Connections conference: “10 Scripts to Make Your Life Easy” for 60+ school technology staff

Certifications

2026
Datadog Fundamentals
2018
Microsoft MCSE: Cloud Platform and Infrastructure, including Azure Stack
2016
Microsoft Specialist: Implementing Azure Infrastructure Solutions
2015
Microsoft MCSA: Office 365
Master Bike Mechanic, Bikes Together
2012
Microsoft MCITP: Enterprise Administrator, plus 16 more Microsoft certifications See allHide
  • MCTS: Windows 7, Configuration (2010)
  • MCITP: Enterprise Desktop Administrator 7 (2010)
  • MCTS: Windows Server 2008 Network Infrastructure, Configuration (2010)
  • MCTS: Windows Server 2008 Active Directory, Configuration (2011)
  • MCITP: Server Administrator on Windows Server 2008 (2011)
  • MCTS: Windows Server 2008 R2, Server Virtualization (2011)
  • MCTS: Windows Server 2008 R2, Desktop Virtualization (2011)
  • MCITP: Virtualization Administrator on Windows Server 2008 R2 (2011)
  • MCTS: Windows Server 2008 Applications Infrastructure, Configuration (2012)
  • MCTS: Administering and Deploying System Center 2012 Configuration Manager (2013)
  • MCTS: System Center Configuration Manager 2007, Configuration
  • MCTS: Microsoft Desktop Optimization Pack, Configuration
  • MCTS: Windows Server Virtualization, Configuration
  • MCTS: Configuring and Deploying a Private Cloud with System Center 2012
  • Microsoft Specialist: Server Virtualization with Windows Server Hyper-V and System Center
  • Microsoft Certified Professional (MCP)
Earlier
Cisco CCNA and CCDA, earned early in my career
CompTIA Security+ and A+

Studied Networking Systems Technology at Linn State Technical College, now State Technical College of Missouri.

The network I run myself

I run my own multi-site network: Home Assistant with 1,000+ IoT devices, the full UniFi stack, and Blue Iris cameras with custom AI detection that protect my shop and my neighbors. It is where I try things before I trust them anywhere else.

Tools I use every day

AI
Claude Code, Google Antigravity, Jev for classification, LLM agents, retrieval and evaluation, MCP, Meilisearch
Platform
Kubernetes and AKS, Azure with Cloud Adoption Framework landing zones and FinOps, Linux and Windows Server, Terraform, Crossplane, Cloudflare Workers and Durable Objects
Delivery
GitOps, GitLab, TeamCity, Octopus Deploy, GitHub Actions
Observability
OpenTelemetry pipelines with Vector, Prometheus, Grafana, OpenSearch, Datadog, Splunk, Azure Monitor, SquaredUp, PostHog
Data
Azure SQL, Cloudflare D1, Power BI pipelines and dashboards
Languages
TypeScript, Python, Go, PowerShell, Bash

For fun

I'm a licensed ham radio operator, and I shoot digital photography and drone video. And of course I repair and ride e-bikes. That’s my happy place.