Leawood, Kansas · Washington, D.C.
I build the systems that decide what to show, what to charge, and what to recommend.
Fifteen years of production machine learning, starting with radar and sensor signal processing at MIT Lincoln Laboratory and running through live events, commerce, luxury retail, and asset management. I co-founded Katch, where we built the content and audience intelligence platform now used by Warner Bros. Discovery, AMC, Meta, and Paramount. I advise the CIO of the U.S. Department of the Treasury on enterprise AI. Chicago Booth MBA, BU master's in electrical and computer engineering, and a Marine before any of it.
- 100k+
- titles in the Katch catalog
- Every one decomposed into machine-comparable attributes.
- $B+
- equity volume scheduled
- GMO's global trade scheduler, which I owned end to end.
- 7
- industries shipped into
- Media, live events, commerce, asset management, health, government, defense.
- 8
- years running Katch
- Blank slate to a profitable business with named enterprise logos.
Currently
U.S. Department of the Treasury
Technical Advisor to the Chief Information Officer, AI Architecture · Mar 2026 to present
Advising the federal CIO on how enterprise AI actually gets deployed, governed, and scaled inside government, and leading the rebuild of a taxpayer app with a nine-figure install base.
Katch
Co-Founder & Chief Technology Officer · Sept 2018 to present
Built a content intelligence company from a blank slate to profitability, and the six products that sit on top of it.
What I have actually worked on
The same handful of problems keep reappearing in industries that think they are nothing alike. A ticket, a title, a storefront visit, and a security are all inventory that has to be matched to a person under a constraint. Here is where I have solved each of them.
| Media | Live events | Retail | Asset mgmt | Health | Government | Defense | Independent | |
|---|---|---|---|---|---|---|---|---|
| Personalization & recommendationMatching a person to a thing when the catalog is large and the signal is thin. | Media & entertainment: yes | Live events: yes | Retail & commerce: yes | Asset management: no | Healthcare: no | Government: no | Defense & research: no | Independent builds: yes |
| Ranking, matching & retrievalOrdering candidates under constraints, and finding the right ones to order in the first place. | Media & entertainment: yes | Live events: no | Retail & commerce: no | Asset management: no | Healthcare: no | Government: yes | Defense & research: no | Independent builds: yes |
| Pricing & market designTwo-sided marketplaces, perishable inventory, and pricing that optimizes lifetime value. | Media & entertainment: no | Live events: yes | Retail & commerce: yes | Asset management: yes | Healthcare: no | Government: no | Defense & research: no | Independent builds: no |
| Time series & signal processingSpectral and time-domain methods, from radar and sensor work at Lincoln Laboratory to buy-side signal construction. | Media & entertainment: no | Live events: no | Retail & commerce: no | Asset management: yes | Healthcare: no | Government: no | Defense & research: yes | Independent builds: yes |
| Visual & creative intelligenceScoring imagery on how it performs with an audience rather than on how it looks, and using the response to it as a signal. | Media & entertainment: yes | Live events: no | Retail & commerce: yes | Asset management: no | Healthcare: no | Government: no | Defense & research: no | Independent builds: no |
| Generative AI & agentsFine-tuned models, retrieval systems, and agents that hand off cleanly to a human. | Media & entertainment: yes | Live events: no | Retail & commerce: no | Asset management: no | Healthcare: yes | Government: yes | Defense & research: no | Independent builds: yes |
| Loyalty, CRM & lifecycleDeciding who to reach, when, and with what, on surfaces the customer is not currently looking at. | Media & entertainment: no | Live events: yes | Retail & commerce: yes | Asset management: no | Healthcare: yes | Government: no | Defense & research: no | Independent builds: no |
| Identity & entity resolutionReconciling the same title, person, or creator across datasets that disagree about all three. | Media & entertainment: yes | Live events: no | Retail & commerce: no | Asset management: no | Healthcare: no | Government: yes | Defense & research: no | Independent builds: yes |
| Experimentation & measurementHoldouts, counterfactuals, and the discipline of proving a model earned its place. | Media & entertainment: yes | Live events: yes | Retail & commerce: yes | Asset management: yes | Healthcare: no | Government: no | Defense & research: no | Independent builds: no |
Where, and when
Katch has run continuously since 2018, underneath everything else. The overlap is deliberate: the operating roles were how I kept testing the same ideas against industries that would not tolerate a demo.
The through-line: personalization
Recommendation is not one problem. It is six or seven, and the interesting ones are rarely the model. These are the pieces I have had to solve in production, and where.
Cold start on a long tail
Katch
Most of a 100,000 title catalog has almost no interaction history. The genome decomposes a title into machine-comparable attributes, so a new title inherits a neighborhood from its content instead of waiting years for behavior to accumulate.
Optimizing inside hard constraints
Ticketmaster / Live Nation
Rights holders, promoters, and venues each set boundaries a model is not permitted to cross. The reinforcement-learning policy optimized inside those limits, which is the difference between a pricing model that is interesting and one that ships.
Reaching people off-platform
Caruso
The hardest surface to personalize is the one nobody is looking at. Luxury retail generates its richest signal in person and captures almost none of it, so the work was capturing that signal and deciding who to contact, when, and with what.
Choosing the artwork, not just the title
Vizit, Katch Party
At Vizit the entire product was predicting how a specific image performs with a specific audience, which is asset selection stated plainly. Katch Party approaches it from the other side, learning taste from poster swipes, so the artwork is both the thing being rated and the thing being personalized.
Explaining the recommendation
Katch
Buyers making eight-figure acquisition decisions do not act on a similarity score. Comparables ship with the attributes that produced them, which changes both adoption and the speed at which a bad model gets caught.
Proving it worked
Ticketmaster, Caruso, GMO
Backtests at GMO, pricing holdouts at Ticketmaster, and lifecycle experiments at Caruso. The through-line is measuring against a counterfactual rather than against the previous quarter.
Katch, in full
One foundation, six products, eight years. Everything Katch ships sits on the same content intelligence layer, which is the reason a creator-vetting tool and a consumer identity passport can come out of the same company without either being a distraction.
The Media Genome
Ingestion & entity resolution
One identity per title, creator, and audience, across providers that disagree about all three.
Semantic decomposition
Content broken into machine-comparable attributes instead of genre labels.
Models & serving
Fine-tuned LLMs, vector search, clustering, and scoring behind one API.
Katch Data↗
katchdata.com
The content intelligence platform. Semantic analysis across film, television, social, music, and books, exposed as entertainment metadata anyone downstream can query.
Live
Katch Verified↗
verified.katchdata.com
Creator vetting for brands and agencies. Scores an influencer's full body of work for brand fit and safety in real time, so a partnership decision stops being a manual review.
Live, account required
Audience Genome↗
audience.katch.ai
Audience profiling and comparable-title analysis across a catalog of more than 100,000 titles. Decomposes a title into machine-comparable attributes so comps are computed instead of argued.
Live, account required
Identity Passport↗
katch.ai
An identity layer for personal AI. A person authors their own profile once and shares it selectively with Claude, ChatGPT, Gemini, and anything else that would otherwise be guessing.
Live
My Katch↗
my.katch.ai
The consumer surface for the Identity Passport: what a person has told us, what each connected service can see, and the controls that govern it.
Live, account required
Katch Party↗
katchparty.com
Tinder for movies. You swipe through posters, and the ratings you give the artwork become a taste profile. It resolves a group's picks into one recommendation everybody can live with, which is the actual problem when two people are deciding what to watch.
Live, account required
In production with
Selected work
Katch
2018 to present
Audience Genome
A recommendation and comparable-title engine over 100,000+ titles, built to work on the long tail where almost no interaction history exists.
Katch
2023 to present
Katch Verified
Real-time brand-safety and brand-fit scoring across a creator's entire body of work, so a partnership decision stops being a manual review.
Katch
2021 to present
Katch Party
Tinder for movies. Rate posters, get recommendations, and resolve a group's competing taste into a single pick everybody can live with.
Ticketmaster / Live Nation
2019 to 2020
Marketplace Pricing & Reinforcement Learning
Reinforcement-learning policies balancing real-time supply and demand across a national two-sided live-entertainment marketplace.
Vizit
2015 to 2018
Image Performance Modeling
Predicting how a specific image performs with a specific audience, so the choice of creative stops being a matter of taste.
GMO LLC
2015 to 2018
Buy-Side Quantitative Research
Signal and valuation research across global equities and fixed income, on the firm-wide research team, alongside the portfolio managers who traded the output.
Caruso
2020 to 2023
Luxury Retail CRM, Loyalty & Dynamic Pricing
Personalization for customers who are nowhere near a screen: capturing offline luxury-retail behavior and using it to decide who to reach, when, and with what.
U.S. Department of the Treasury
2026 to present
IRS2Go Revamp
Leading the rebuild of the IRS mobile app, the single consumer product touching essentially every U.S. taxpayer.
U.S. Department of the Treasury
2026 to present
Enterprise AI Architecture Advisory
Advising the federal CIO on deployment, risk assessment, and scaling of enterprise AI across government platforms.
Also
Identity Passport
Katch
An identity layer for personal AI. Author your profile once, share it selectively, and stop letting every assistant guess independently.
Global Equity Trade Scheduler
GMO LLC
Replaced the spreadsheet workflow behind billions of dollars of equity execution with an automated, error-resistant platform.
AI Clinical Intake Agent
Sully.ai
An AI agent handling routine clinical intake, plus the CRM, scheduling, and handoff layer that makes it safe to pass a patient to a human.
Radar & Sensor Signal Processing
MIT Lincoln Laboratory
Detection and estimation against noisy, nonstationary sensor data, and the origin of everything I think about signal.
Koobrik
Independent
A serverless tool that reads a comic book page by page and synthesizes it into studio-style coverage.
Gigs
Independent / Consulting
A two-sided job marketplace: seeker app, employer app, and the API, identity, and billing backbone underneath both.
Experience
Mar 2026 to present
Washington, D.C.
U.S. Department of the Treasury
Technical Advisor to the Chief Information Officer, AI Architecture
- Advise the federal CIO and senior leadership on deployment, risk assessment, and scaling of enterprise AI architecture across government platforms, covering LLMs, retrieval-augmented generation, and agentic systems.
- Lead the revamp of IRS2Go, the IRS mobile app, working across product, security, and accessibility constraints that apply to every U.S. taxpayer.
- Presented AI architecture evaluations and deployment frameworks to senior officials.
- Presented at the White House on managing AI engineering teams, and on what has to change in the product function when engineering output rises faster than the process around it can absorb.
- Assess emerging AI platforms and vendors for security, compliance, and scalability to guide federal integration strategy.
Sept 2018 to present
Burbank, CA
Katch
Co-Founder & Chief Technology Officer
- Founded and scaled a media-data analytics company from nothing to profitability, iterating through discovery to product-market fit.
- Architected the Media Genome: ingestion, entity resolution, and semantic decomposition of film, television, social, music, and books into structured, machine-comparable attributes.
- Shipped six products on that foundation, spanning entertainment metadata, creator verification and brand safety, audience matching across 100,000+ titles, and a consumer identity layer for personal AI.
- Fine-tuned LLMs for content understanding and for turning natural-language questions into the analytical queries underneath them.
- Lead engineering and product, and run technical evaluation of partnerships and acquisition targets.
Oct 2025 to Jan 2026
Remote
Sully.ai
Head of Product
- Led product vision for an AI clinical-intake agent that automates roughly 80% of routine tasks, freeing staff for high-value patient care.
- Built a workflow-heavy CRM and scheduling platform integrated with legacy EMRs, and owned the AI-to-human handoff experience.
- Shipped internal tools that let non-technical staff configure AI behavior, cutting new-practice onboarding time by half.
Aug 2020 to Mar 2023
Beverly Hills, CA
Caruso
Director of Data & Analytics
- Led product strategy for CRM and loyalty across a luxury retail portfolio, digitizing in-person customer interactions into structured, addressable data.
- Partnered with executive leadership to deploy dynamic-pricing models optimized for long-term customer lifetime value rather than the next transaction.
- Built the segmentation and lifecycle logic behind outbound offers, which is personalization for customers who are nowhere near a screen.
July 2019 to May 2020
Hollywood, CA
Ticketmaster / Live Nation
Senior Product Data Scientist
- Owned pricing products for a national two-sided live-entertainment marketplace, from the model through the levers promoters and venues actually touch.
- Deployed reinforcement-learning policies to balance real-time supply and demand inside constraints set by rights holders.
- Instrumented outcomes so pricing decisions were evaluated against a counterfactual instead of against intuition.
Aug 2015 to July 2018
Boston, MA
GMO LLC
Quantitative Researcher & Developer, Centralized Research Group
- Quantitative researcher and developer in the firm-wide research group covering global equities and fixed income, building signal and valuation models directly with the portfolio managers who traded them.
- Worked the full research loop: hypothesis, data construction, backtest, statistical validation, and the production code that ran the result.
- Product-owned the Global Equity Trade Scheduler, the mission-critical system executing billions of dollars in volume, replacing manual spreadsheet workflows with an automated, error-resistant platform.
2015 to 2018
Remote
Vizit
Principal Data Scientist
- Principal data scientist at a computer-vision startup founded by friends, scoring product and brand imagery for how well it performs with a specific audience rather than for whether it looks good.
- The problem is asset selection: same product, several images, and the question of which one to put in front of which person. It is the same shape as the artwork question in a recommendation feed.
- Held alongside the GMO role, with the Booth MBA running over much of the same period.
2012 to 2015
Lexington, MA
MIT Lincoln Laboratory
Technical Staff
- Technical staff on radar and sensor signal processing: detection and estimation against noisy, nonstationary data where the cost of a false positive is not symmetric with the cost of a miss.
- Spectral and time-domain methods applied to real instrument data rather than to textbook signals, which is where you learn how much apparent structure is an artifact of the measurement.
- Finished the BU master's in electrical and computer engineering across the same three years.
Summer 2016
Austin, TX (Remote)
G51
Venture Capital Summer Associate
- Sourced and evaluated early-stage technology companies, running financial and technical diligence, market sizing, and competitive analysis for the investment team.
Labs
Smaller builds and experiments. Some shipped, some deliberately abandoned once they answered the question that prompted them.
TTB High-Velocity Compliance Workbench↗
A compliance review workbench built for high-throughput regulatory audit workflows.
TypeScript
DME Voice Agent↗
A voice agent for durable medical equipment intake and qualification calls.
Python
Mirror Mirror
Virtual try-on as a Chrome extension. Transfers a garment from any product page onto your own photo, with an agentic quality-control loop.
Python, FastAPI, Gemini
Fine-Tuning as a Service
Production-grade LLM fine-tuning platform: SFT and DPO pipelines, vLLM serving, MCP tooling, and enterprise observability.
Python, vLLM, MCP
Here-o
Crowdsourced personal safety, connecting people to a nearby network that has their back.
TypeScript, React Native
Handwritten Calendar to iCal
Photograph a handwritten wall calendar, get back a working .ics file.
Python, computer vision
ARC Prize 2025
Program-synthesis approaches to the Abstraction and Reasoning Corpus, the standing benchmark for fluid intelligence in machines.
Python
Brick Studio
Converts any photograph into an official-palette LEGO mosaic and renders step-by-step build instructions.
Python, image processing
Capabilities
Machine learning
- Recommendation & personalization
- Ranking and retrieval
- Reinforcement learning
- LLM fine-tuning (SFT, DPO)
- Retrieval-augmented generation
- Agentic systems
- Experimentation & causal measurement
Data & engineering
- Python
- C++
- SQL
- TypeScript
- Entity resolution
- Distributed data pipelines
- AWS & GCP
Research & strategy
- Quantitative research
- Time-series & spectral methods
- Pricing & market design
- Technical diligence
- Product strategy
- Engineering leadership
Building things that have to work, and advising the people funding them.
I take on a small number of engagements at a time: personalization and ranking systems, AI and data architecture, and technical diligence on AI and media businesses. If that overlaps with what you need, get in touch.