sean kearns
HomeLet's Build.

Flight log / Marquette → now

My
Timeline.

Going from school to work.

Explore my GitHub

01 / The main timeline

Rewind the work.

From modeling the real world at Marquette to building the data foundations and agents that put those ideas to work.

2026PRESENT DAY

February 2026 — Present

Purveyor Sports

Data & AI Engineering

I bring sports data together into persistent fan profiles, then build agents that can work with that context. My work spans data infrastructure, identity resolution, and Sixth Man: an AI revenue agent with stateful, multi-touch workflows.

Explore the work ↗Related writing: Using Event-Level Data to Make Hyper-Personalized LLMs ↗
2025ARCHIVE / 02

March 2025 — February 2026

Chicago Fire FC

Sr. Manager, Business Strategy and Analytics

I led a team across analytics, data science, and CRM, bringing predictive modeling into ticket retention and revenue decisions. I also led AI adoption, tested LLM workflows with leadership, and deployed Conversica to support stadium lead qualification.

Explore the work ↗Related writing: Using Event-Level Data to Make Hyper-Personalized LLMs ↗
2022ARCHIVE / 03

January 2022 — March 2025

Chicago Fire FC

Full Stack Data Engineer

I built the foundation: data ingestion, a central warehouse, and a front-office portal. React and Django connected the business to Snowflake; automated pipelines brought ticketing, CRM, and engagement data into the same picture.

Explore the work ↗Related writing: Building a Sovereign Tech Stack in Sports ↗
BeginningsEDUCATION / ORIGIN STORY

Where the questions started

Marquette University

Mathematics & Economics

At Marquette, I studied mathematics and economics and became obsessed with how models could explain the real world. But applying what I was learning required good data—and good data was hard to find.

The idea I carried forward

That realization drew me toward data engineering. I wanted to build impactful solutions for aggregating, consolidating, and cleaning data, giving new applications a reliable foundation to build on.

Models → Better data → Useful applications

02 / Parallel paths

Companies & collaborations.

Consulting, advisory work, and partnerships alongside the main timeline.

During my Purveyor work

Momento & Venue Vision

Partner work through Purveyor Sports

Computer vision and data engineering for venue intelligence: transforming stadium imagery into structured datasets using YOLOv8 and MiVOLO, and connecting that work to the broader fan-data platform.

Explore the work ↗

September 2025 — Present

SpikeHat Games

Engineering Consultant

Website design and development for a board game company, including game pages and blog posts, with work on digital strategy, go-to-market planning, and integrated payments.

Explore the work ↗

February 2025 — June 2025

Hale

Advisor

Advised a product recommendation service. Helped establish Firebase deployments and CI/CD, built a Supabase database, and connected Mixpanel tracking to understand how people interact with products.

Explore the work ↗

October 2024 — February 2025

Revenue Base

Engineering Consultant

Built Domo data pipelines and dashboards for deal metrics, lead growth, and renewal health.

Explore the work ↗

03 / After hours

Things I wanted to exist.

Personal projects are part of the story, too.

November 2024 — Present

Handle

Personal project

A platform concept for bringing personal data into one place, with users in control of what they own, share, and sell. It combines health, nutrition, activity, and daily context with linked, verified social profiles—creating useful applications and a marketplace where users can choose to grant access to their data and receive payment. Marketplace rates and payouts are currently simulated.

Explore the work ↗Project repository ↗

The systems behind the work

Tools I use.

A few familiar companions, always in motion. Select a tool to see how I use it.

Explore the rest of my toolkit

Data foundations

Snowflake

Cloud data warehouse

Where I use it

Centralizing ticketing, CRM, and engagement data at the Fire and Purveyor.

Next experiments

Proposed experiment: make source quality, transformations, and identity resolution visible alongside every downstream insight.

Working hypothesis: a well-modeled data layer becomes more valuable as more applications and agents share it.

Cloud & delivery

AWS

Cloud infrastructure & delivery

Where I use it

Compute, storage, deployment, and monitoring across my data and application work.

Agents & AI

LangChain

Model and tool integrations

Where I use it

Sixth Man agent development at Purveyor.

Explore the repository
Next experiments

Proposed experiment: compare stateful workflows against simpler baselines, with evaluation and human review before consequential actions.

Working hypothesis: useful agents will depend as much on trustworthy context and observable execution as on the model itself.

Data foundations

Prefect

Workflow orchestration

Where I use it

Automated ingestion and data pipelines at the Fire and Purveyor.

Explore the repository
Next experiments

Proposed experiment: make source quality, transformations, and identity resolution visible alongside every downstream insight.

Working hypothesis: a well-modeled data layer becomes more valuable as more applications and agents share it.

Agents & AI

LangGraph

Stateful agent orchestration

Where I use it

Multi-touch workflows that incorporate evolving fan context at Purveyor.

Explore the repository
Next experiments

Proposed experiment: compare stateful workflows against simpler baselines, with evaluation and human review before consequential actions.

Working hypothesis: useful agents will depend as much on trustworthy context and observable execution as on the model itself.

Data foundations

dbt

Versioned data transformations

Where I use it

Data modeling in the sports data stack.

Explore the repository
Next experiments

Proposed experiment: make source quality, transformations, and identity resolution visible alongside every downstream insight.

Working hypothesis: a well-modeled data layer becomes more valuable as more applications and agents share it.

Data foundations

Python

Data and application programming

Where I use it

Pipelines, automation, APIs, and agent workflows.

Next experiments

Proposed experiment: make source quality, transformations, and identity resolution visible alongside every downstream insight.

Working hypothesis: a well-modeled data layer becomes more valuable as more applications and agents share it.

Apps & integrations

React

Interactive interfaces

Where I use it

Fire front-office portal and Handle.

Next experiments

Proposed experiment: connect more user-controlled signals to useful interfaces, with clear permissions and visible provenance.

Working hypothesis: the best interface helps people understand and act on their context, rather than just collecting more of it.

Cloud & delivery

Docker

Application containers

Where I use it

Packaging ingestion workflows.

Next experiments

Proposed experiment: package repeatable deployments with explicit costs, observable failures, and smaller operational footprints.

Working hypothesis: infrastructure should make experimentation easier without hiding how a system runs.

Apps & integrations

Supabase

Application database platform

Where I use it

Handle’s health model and Hale’s product database.

Next experiments

Proposed experiment: connect more user-controlled signals to useful interfaces, with clear permissions and visible provenance.

Working hypothesis: the best interface helps people understand and act on their context, rather than just collecting more of it.

Agents & AI

OpenAI API

Language-model applications

Where I use it

Fan outreach, summarization, text-to-SQL, and health-context summaries.

Next experiments

Proposed experiment: compare stateful workflows against simpler baselines, with evaluation and human review before consequential actions.

Working hypothesis: useful agents will depend as much on trustworthy context and observable execution as on the model itself.

Data foundations

PostgreSQL

Relational database

Where I use it

Ticketing data streams and application data.

Next experiments

Proposed experiment: make source quality, transformations, and identity resolution visible alongside every downstream insight.

Working hypothesis: a well-modeled data layer becomes more valuable as more applications and agents share it.

Agents & AI

Amazon Bedrock AgentCore

Infrastructure for running and connecting agents

Where I use it

Part of my recent agent toolbelt; a detailed project note is still to come.

Official documentation
Next experiments

Proposed experiment: compare stateful workflows against simpler baselines, with evaluation and human review before consequential actions.

Working hypothesis: useful agents will depend as much on trustworthy context and observable execution as on the model itself.

Agents & AI

Amazon Bedrock

Managed foundation-model access

Where I use it

Model access in the Sixth Man stack.

Next experiments

Proposed experiment: compare stateful workflows against simpler baselines, with evaluation and human review before consequential actions.

Working hypothesis: useful agents will depend as much on trustworthy context and observable execution as on the model itself.

Agents & AI

Snowflake Cortex

AI capabilities alongside warehouse data

Where I use it

AI workflows over customer context at Purveyor.

Next experiments

Proposed experiment: compare stateful workflows against simpler baselines, with evaluation and human review before consequential actions.

Working hypothesis: useful agents will depend as much on trustworthy context and observable execution as on the model itself.

Agents & AI

Conversica

Conversational sales automation

Where I use it

Lead qualification for the Chicago Fire stadium go-to-market effort.

Next experiments

Proposed experiment: compare stateful workflows against simpler baselines, with evaluation and human review before consequential actions.

Working hypothesis: useful agents will depend as much on trustworthy context and observable execution as on the model itself.

Agents & AI

YOLOv8

Object detection

Where I use it

Stadium imagery workflows with Momento through Purveyor.

Explore the repository
Next experiments

Proposed experiment: compare stateful workflows against simpler baselines, with evaluation and human review before consequential actions.

Working hypothesis: useful agents will depend as much on trustworthy context and observable execution as on the model itself.

Agents & AI

MiVOLO

Visual attribute estimation

Where I use it

Computer vision pipelines with Momento through Purveyor.

Explore the repository
Next experiments

Proposed experiment: compare stateful workflows against simpler baselines, with evaluation and human review before consequential actions.

Working hypothesis: useful agents will depend as much on trustworthy context and observable execution as on the model itself.

Agents & AI

AWS SageMaker

Machine-learning tooling

Where I use it

Listed in my Purveyor infrastructure tool history; project detail to come.

Next experiments

Proposed experiment: compare stateful workflows against simpler baselines, with evaluation and human review before consequential actions.

Working hypothesis: useful agents will depend as much on trustworthy context and observable execution as on the model itself.

Data foundations

SQL

Querying and modeling data

Where I use it

Warehouse transformations, investigation, and analytics.

Next experiments

Proposed experiment: make source quality, transformations, and identity resolution visible alongside every downstream insight.

Working hypothesis: a well-modeled data layer becomes more valuable as more applications and agents share it.

Data foundations

R

Statistical analysis

Where I use it

Part of my Chicago Fire analytics toolkit.

Next experiments

Proposed experiment: make source quality, transformations, and identity resolution visible alongside every downstream insight.

Working hypothesis: a well-modeled data layer becomes more valuable as more applications and agents share it.

Data foundations

Fivetran

Managed data connectors

Where I use it

Connector-related API work at Boost Sports.

Next experiments

Proposed experiment: make source quality, transformations, and identity resolution visible alongside every downstream insight.

Working hypothesis: a well-modeled data layer becomes more valuable as more applications and agents share it.

Data foundations

Domo

Data pipelines and dashboards

Where I use it

Deal, lead, and renewal reporting for Revenue Base.

Next experiments

Proposed experiment: make source quality, transformations, and identity resolution visible alongside every downstream insight.

Working hypothesis: a well-modeled data layer becomes more valuable as more applications and agents share it.

Data foundations

Sigma

Warehouse-connected analytics

Where I use it

Business analytics in the sports data stack.

Next experiments

Proposed experiment: make source quality, transformations, and identity resolution visible alongside every downstream insight.

Working hypothesis: a well-modeled data layer becomes more valuable as more applications and agents share it.

Data foundations

Tableau

Visual analysis and dashboards

Where I use it

Chicago Fire business reporting.

Next experiments

Proposed experiment: make source quality, transformations, and identity resolution visible alongside every downstream insight.

Working hypothesis: a well-modeled data layer becomes more valuable as more applications and agents share it.

Data foundations

Google Sheets

Collaborative spreadsheets

Where I use it

Source workflows before Dyer Brothers’ CRM migration.

Next experiments

Proposed experiment: make source quality, transformations, and identity resolution visible alongside every downstream insight.

Working hypothesis: a well-modeled data layer becomes more valuable as more applications and agents share it.

Data foundations

Google Colab

Hosted notebooks

Where I use it

Listed in my Dyer Brothers tool history; notebook detail to come.

Next experiments

Proposed experiment: make source quality, transformations, and identity resolution visible alongside every downstream insight.

Working hypothesis: a well-modeled data layer becomes more valuable as more applications and agents share it.

Cloud & delivery

AWS S3

Object storage

Where I use it

Sports data ingestion and application assets.

Next experiments

Proposed experiment: package repeatable deployments with explicit costs, observable failures, and smaller operational footprints.

Working hypothesis: infrastructure should make experimentation easier without hiding how a system runs.

Cloud & delivery

AWS Lambda

Event-driven functions

Where I use it

Automated cleansing and integration work.

Next experiments

Proposed experiment: package repeatable deployments with explicit costs, observable failures, and smaller operational footprints.

Working hypothesis: infrastructure should make experimentation easier without hiding how a system runs.

Cloud & delivery

AWS EC2

Compute instances

Where I use it

Data services and the NYT leaderboard deployment.

Next experiments

Proposed experiment: package repeatable deployments with explicit costs, observable failures, and smaller operational footprints.

Working hypothesis: infrastructure should make experimentation easier without hiding how a system runs.

Cloud & delivery

AWS CloudWatch

Logs and monitoring

Where I use it

Operating data pipelines and hosted applications.

Next experiments

Proposed experiment: package repeatable deployments with explicit costs, observable failures, and smaller operational footprints.

Working hypothesis: infrastructure should make experimentation easier without hiding how a system runs.

Cloud & delivery

AWS CloudFront

Content delivery

Where I use it

Listed in the Fire infrastructure inventory.

Next experiments

Proposed experiment: package repeatable deployments with explicit costs, observable failures, and smaller operational footprints.

Working hypothesis: infrastructure should make experimentation easier without hiding how a system runs.

Cloud & delivery

AWS Route 53

DNS management

Where I use it

Hosting the Fire front-office portal.

Next experiments

Proposed experiment: package repeatable deployments with explicit costs, observable failures, and smaller operational footprints.

Working hypothesis: infrastructure should make experimentation easier without hiding how a system runs.

Cloud & delivery

AWS ECR

Container registry

Where I use it

Containerized ingestion at the Fire.

Next experiments

Proposed experiment: package repeatable deployments with explicit costs, observable failures, and smaller operational footprints.

Working hypothesis: infrastructure should make experimentation easier without hiding how a system runs.

Cloud & delivery

AWS Fargate

Managed container execution

Where I use it

Prefect pipeline infrastructure at the Fire.

Next experiments

Proposed experiment: package repeatable deployments with explicit costs, observable failures, and smaller operational footprints.

Working hypothesis: infrastructure should make experimentation easier without hiding how a system runs.

Cloud & delivery

AWS Pinpoint

Communications tooling

Where I use it

Phone-number analysis for Dyer Brothers.

Next experiments

Proposed experiment: package repeatable deployments with explicit costs, observable failures, and smaller operational footprints.

Working hypothesis: infrastructure should make experimentation easier without hiding how a system runs.

Cloud & delivery

Kubernetes

Container orchestration

Where I use it

Automated ingestion infrastructure at the Fire.

Next experiments

Proposed experiment: package repeatable deployments with explicit costs, observable failures, and smaller operational footprints.

Working hypothesis: infrastructure should make experimentation easier without hiding how a system runs.

Cloud & delivery

Terraform

Infrastructure as code

Where I use it

Listed in my Boost Sports toolkit.

Next experiments

Proposed experiment: package repeatable deployments with explicit costs, observable failures, and smaller operational footprints.

Working hypothesis: infrastructure should make experimentation easier without hiding how a system runs.

Cloud & delivery

Nginx

Web serving and reverse proxy

Where I use it

Fire portal and NYT leaderboard hosting.

Next experiments

Proposed experiment: package repeatable deployments with explicit costs, observable failures, and smaller operational footprints.

Working hypothesis: infrastructure should make experimentation easier without hiding how a system runs.

Cloud & delivery

Gunicorn

Python web application serving

Where I use it

NYT leaderboard deployment.

Next experiments

Proposed experiment: package repeatable deployments with explicit costs, observable failures, and smaller operational footprints.

Working hypothesis: infrastructure should make experimentation easier without hiding how a system runs.

Cloud & delivery

Vercel

Web application hosting

Where I use it

Handle, consulting websites, and personal projects.

Next experiments

Proposed experiment: package repeatable deployments with explicit costs, observable failures, and smaller operational footprints.

Working hypothesis: infrastructure should make experimentation easier without hiding how a system runs.

Cloud & delivery

Firebase

Application platform

Where I use it

Deployment and CI/CD work for Hale.

Next experiments

Proposed experiment: package repeatable deployments with explicit costs, observable failures, and smaller operational footprints.

Working hypothesis: infrastructure should make experimentation easier without hiding how a system runs.

Cloud & delivery

Heroku

Application hosting

Where I use it

Earlier ticketing data-streaming work at the Fire.

Next experiments

Proposed experiment: package repeatable deployments with explicit costs, observable failures, and smaller operational footprints.

Working hypothesis: infrastructure should make experimentation easier without hiding how a system runs.

Cloud & delivery

GoDaddy

Domain tooling

Where I use it

Listed in the original NYT leaderboard toolkit.

Next experiments

Proposed experiment: package repeatable deployments with explicit costs, observable failures, and smaller operational footprints.

Working hypothesis: infrastructure should make experimentation easier without hiding how a system runs.

Cloud & delivery

GitHub

Source control and collaboration

Where I use it

Repositories and delivery across my projects.

Next experiments

Proposed experiment: package repeatable deployments with explicit costs, observable failures, and smaller operational footprints.

Working hypothesis: infrastructure should make experimentation easier without hiding how a system runs.

Cloud & delivery

Jira

Work tracking

Where I use it

Listed in my Fire and Boost Sports tool history.

Next experiments

Proposed experiment: package repeatable deployments with explicit costs, observable failures, and smaller operational footprints.

Working hypothesis: infrastructure should make experimentation easier without hiding how a system runs.

Apps & integrations

JavaScript

Web programming

Where I use it

Websites, interfaces, and application integrations.

Next experiments

Proposed experiment: connect more user-controlled signals to useful interfaces, with clear permissions and visible provenance.

Working hypothesis: the best interface helps people understand and act on their context, rather than just collecting more of it.

Apps & integrations

Django

Python web applications

Where I use it

Fire portal backend and NYT leaderboard.

Next experiments

Proposed experiment: connect more user-controlled signals to useful interfaces, with clear permissions and visible provenance.

Working hypothesis: the best interface helps people understand and act on their context, rather than just collecting more of it.

Apps & integrations

Salesforce

Customer relationship management

Where I use it

Connecting fan records and activation at the Fire.

Next experiments

Proposed experiment: connect more user-controlled signals to useful interfaces, with clear permissions and visible provenance.

Working hypothesis: the best interface helps people understand and act on their context, rather than just collecting more of it.

Apps & integrations

HubSpot

Customer relationship management

Where I use it

CRM migration for Dyer Brothers.

Next experiments

Proposed experiment: connect more user-controlled signals to useful interfaces, with clear permissions and visible provenance.

Working hypothesis: the best interface helps people understand and act on their context, rather than just collecting more of it.

Apps & integrations

Ticketmaster / Archtics

Ticketing data source

Where I use it

Ticket history and near-real-time streams at the Fire.

Next experiments

Proposed experiment: connect more user-controlled signals to useful interfaces, with clear permissions and visible provenance.

Working hypothesis: the best interface helps people understand and act on their context, rather than just collecting more of it.

Apps & integrations

Okta

Identity and access

Where I use it

Listed in the Fire application toolkit.

Next experiments

Proposed experiment: connect more user-controlled signals to useful interfaces, with clear permissions and visible provenance.

Working hypothesis: the best interface helps people understand and act on their context, rather than just collecting more of it.

Apps & integrations

Google OAuth

User authentication

Where I use it

Sign-in for Handle and NYT Leaderboards.

Next experiments

Proposed experiment: connect more user-controlled signals to useful interfaces, with clear permissions and visible provenance.

Working hypothesis: the best interface helps people understand and act on their context, rather than just collecting more of it.

Apps & integrations

Google Analytics

Web measurement

Where I use it

Understanding usage of personal apps and digital properties.

Next experiments

Proposed experiment: connect more user-controlled signals to useful interfaces, with clear permissions and visible provenance.

Working hypothesis: the best interface helps people understand and act on their context, rather than just collecting more of it.

Apps & integrations

Mixpanel

Product analytics

Where I use it

User and product interactions at Hale.

Next experiments

Proposed experiment: connect more user-controlled signals to useful interfaces, with clear permissions and visible provenance.

Working hypothesis: the best interface helps people understand and act on their context, rather than just collecting more of it.

Apps & integrations

WHOOP

Health and recovery data

Where I use it

Connected health context in Handle.

Next experiments

Proposed experiment: connect more user-controlled signals to useful interfaces, with clear permissions and visible provenance.

Working hypothesis: the best interface helps people understand and act on their context, rather than just collecting more of it.

Apps & integrations

Open Food Facts

Food and nutrition data

Where I use it

Food tracking in Handle.

Next experiments

Proposed experiment: connect more user-controlled signals to useful interfaces, with clear permissions and visible provenance.

Working hypothesis: the best interface helps people understand and act on their context, rather than just collecting more of it.

Apps & integrations

Twitter / X

Social data source

Where I use it

Listed in the original Fire source inventory.

Next experiments

Proposed experiment: connect more user-controlled signals to useful interfaces, with clear permissions and visible provenance.

Working hypothesis: the best interface helps people understand and act on their context, rather than just collecting more of it.

Apps & integrations

Facebook

Social data source

Where I use it

Listed in the original Fire source inventory.

Next experiments

Proposed experiment: connect more user-controlled signals to useful interfaces, with clear permissions and visible provenance.

Working hypothesis: the best interface helps people understand and act on their context, rather than just collecting more of it.

Apps & integrations

TikTok

Social data source

Where I use it

Listed in the original Fire source inventory.

Next experiments

Proposed experiment: connect more user-controlled signals to useful interfaces, with clear permissions and visible provenance.

Working hypothesis: the best interface helps people understand and act on their context, rather than just collecting more of it.

Apps & integrations

ZoomInfo

Business data source

Where I use it

Listed in the original Fire source inventory.

Next experiments

Proposed experiment: connect more user-controlled signals to useful interfaces, with clear permissions and visible provenance.

Working hypothesis: the best interface helps people understand and act on their context, rather than just collecting more of it.

Apps & integrations

Marketing Cloud

Marketing data and activation

Where I use it

Listed in the original Fire source inventory.

Next experiments

Proposed experiment: connect more user-controlled signals to useful interfaces, with clear permissions and visible provenance.

Working hypothesis: the best interface helps people understand and act on their context, rather than just collecting more of it.

Apps & integrations

Blinkfire

Engagement measurement

Where I use it

Listed in the original Fire source inventory.

Next experiments

Proposed experiment: connect more user-controlled signals to useful interfaces, with clear permissions and visible provenance.

Working hypothesis: the best interface helps people understand and act on their context, rather than just collecting more of it.

Away from the keyboard

My dashboard.

Films, music, miles, and books. A little of the context behind me.

These live widgets are provided by their platforms. If one is unavailable or asks you to sign in, use its direct link.