Available for Select Engagements

I build the thing, then I run it.

I'm Tue Pham. For fifteen years I've worked the buy side of insurance and performance marketing: acquiring leads, pricing them, routing them, and building the models that decide which ones are worth buying.

Two companies co-founded as an operating executive. Before that, distribution and growth leadership at Fidelity Life, Amazon, Expedia, and a mobile network of 150+ apps. BurntOrange LLC is how I take that work on for other people.

15 yrs
Insurance distribution, lead acquisition, and growth
2x
Co-founder, both times as an operating executive
4
Production scoring models shipped and running

What I Do

Four areas, and they overlap on purpose

Most of the useful work I've done sits where these cross: an insurance product that only works because of the model underneath it, a marketing budget that only improves once someone rebuilds the routing. I take engagements in any one of them, and the others tend to come along.

01

Technology & Applied AI

I write the code and own the architecture, so the advice comes with a working version attached.

  • Predictive scoring models: feature engineering, training, evaluation on AUC, lift, and calibration
  • LLM and agentic products, including autonomous agents that act on live business data
  • Google Cloud builds end to end: BigQuery, Cloud Run, Vertex AI, Firebase, global load balancing
  • Low-latency scoring APIs, real-time decisioning, and the data pipelines underneath them
  • Cloud consumption economics, which is usually where the surprise bills are hiding
02

Insurance Product

Fifteen years inside insurance distribution, mostly on the side that writes the checks.

  • Medicare, ACA, final expense, term life, and property & casualty product strategy
  • Scoring and data products: what to build, what it is worth, and how to prove it
  • Pricing and packaging, including usage-based and value-linked models
  • Carrier and partner co-development, embedding scoring into someone else's workflow
  • Compliance-aware data handling, HIPAA and TCPA constraints shaping the design
03

Growth & Marketing

Acquisition at volume, judged on unit economics rather than impressions.

  • Lead acquisition and buying strategy across digital, call center, and partner channels
  • Performance marketing run against CPA, LTV, and ROAS targets
  • Lead routing, distribution, and vendor onboarding at scale
  • Experimentation and multivariate testing programs, built and run
  • Positioning, messaging, launch, and the sales enablement that follows
04

Operations & COO Work

The unglamorous half: budgets, vendors, reporting, and the meetings where decisions actually get made.

  • Unit economics and the operating model, built so the numbers reconcile
  • Multi-million-dollar budget ownership across internal teams and outside vendors
  • Vendor and partner management, from contracting through performance reviews
  • Cross-functional alignment between marketing, operations, technology, and sales
  • Financial planning, multi-year models, cap table and dilution scenarios

How It Works

Three ways to work together

I keep a small number of engagements at a time, which is the only honest way to do this well. If you are not sure which of these fits, that is a normal place to start the conversation from.

Fractional COO / CPO

Ongoing, part time

An operating executive without the full-time hire. I run a function, sit in the leadership meetings, and own outcomes rather than recommendations.

Usually a fit when

  • Founder-led company that has outgrown the founder running everything
  • Agency, FMO, or lead-gen business that needs product and operating discipline
  • Team with a working product and no one owning the economics

Advisory

Retainer or standing hours

Applied AI, data, and scoring questions answered by someone who has shipped them. Architecture review, build-versus-buy, model evaluation, vendor diligence.

Usually a fit when

  • You are about to spend real money on a model or a data vendor
  • Your team is building something you cannot technically evaluate
  • You want a second opinion that has run the cloud bill before

Project Work

Scoped, fixed deliverable

A defined thing, built and handed over. Often this is the right first engagement, because it proves the working relationship on something small.

Usually a fit when

  • A retrospective study on your historical spend and outcomes
  • A scoring model, routing layer, or decisioning API built and deployed
  • Pricing and packaging work, or a go-to-market plan with the math behind it

Selected Work

What I've actually shipped

The numbers below come from the engagements themselves. Where a result was measured against historical data rather than a live holdout, it says so. The difference matters, and most people selling this kind of work will not tell you which one they are quoting.

RiskLync

Applied AI / Insurance

Co-Founder & Chief Product Officer · 2023 to present

The problem

Health insurance buyers spend billions on leads and cannot tell a good one from a bad one until after they have already paid for it.

What I did

Owned product strategy, pricing, and go-to-market for a scoring platform and a real-time bidding marketplace. Shipped four production models, directed the auction build, and ran the cloud architecture hands on.

88x
Conversion differential, best decile to worst, on a pre-bid model scoring in single-digit milliseconds
<200ms
Ping to bid, end to end, across scoring, campaign filtering, and auction
0.78
AUC on the strongest post-acquisition model in the suite

Worth saying plainly: A retrospective study across 131,119 matched leads and $4.8M of a partner's historical spend showed a 99% lift in Medicare Advantage enrollments at the same budget, and roughly half the cost per acquisition. That is validation against past data, which is the honest way to describe it.

Property & Casualty Scoring

Data Engineering

Built at RiskLync · 2025 to 2026

The problem

Homeowners and auto lead buyers price every ZIP code the same way, even though hazard exposure, home values, and claim history are nothing alike across them.

What I did

Built a pre-bid risk score for property & casualty leads out of federal and public data: FEMA National Risk Index, Zillow home values, IRS income, NOAA storm events, NFIP flood claims, FBI crime, and Census demographics. Precomputed so the lookup lands inside the ping.

82
Features engineered across seven federal and public data sources
37,000+
ZIP profiles precomputed and served from memory
<1ms
Score returned per request

Worth saying plainly: Described here as engineering rather than predictive performance, on purpose. The composite is heuristic until it has been trained against enough outcome data to claim otherwise.

AveriSana

Predictive Analytics

Co-Founder & Chief Operating Officer · 2021 to 2023

The problem

Carriers and agencies had plenty of data on their applicants and no way to act on it before the acquisition cost was already sunk.

What I did

Took a predictive-analytics product from concept to market across Medicare, final expense, and term life. Led customer discovery, co-developed with carriers, and embedded scoring directly into partner acquisition workflows. Owned commercialization and pricing.

~70%
Reduction in one insurtech partner's term-life customer acquisition cost
2
Products taken from prototype to sellable roadmap
3
Insurance lines covered by the scoring suite

eFinancial / Fidelity Life

Distribution

Director, Head of Lead Management · 2017 to 2021

The problem

A direct-to-consumer life insurance business buying in a saturated market, with a vendor onboarding process slow enough to cost it volume.

What I did

Owned acquisition and distribution strategy across internal teams and outside partners. Architected and launched a lead-hosting and routing platform with rapid vendor onboarding and real-time routing, then ran the unit economics across digital and call-center channels.

+20%
Acquisition volume, in a market everyone else called saturated
$M+
Annual budget owned, measured on CPA and revenue per marketing dollar
Real time
Lead routing, replacing a batch handoff

Influence Mobile

Growth at Scale

Vice President, Growth & Marketing · 2015 to 2017

The problem

A small, high-churn user base across a large app portfolio, with acquisition spend that could not be managed by hand.

What I did

Built and coached the marketing team across acquisition, retention, and monetization. Built API-driven systems running tens of thousands of campaigns against ROAS targets, and rebuilt onboarding on the back of structured experiments.

150+
Apps and games in the portfolio, scaled to millions of retained users
+75%
Day-zero ARPU after the onboarding redesign
+50%
Day-one retention from the same work

Background

Where the experience comes from

Fifteen years on the buy side of acquisition, starting in experimentation and games, then into insurance distribution, then twice into founding and running companies that score the leads everyone else is buying blind.

Roles

  1. 2023 to present

    Co-Founder & Chief Product Officer

    RiskLync Inc.

    Applied-AI lead scoring and real-time bidding for the insurance lead market.

  2. 2021 to 2023

    Co-Founder & Chief Operating Officer

    AveriSana Inc.

    Predictive analytics and lead scoring for carriers and agencies.

  3. 2017 to 2021

    Director, Head of Lead Management

    eFinancial / Fidelity Life

    Direct-to-consumer insurance acquisition and distribution.

  4. 2015 to 2017

    Vice President, Growth & Marketing

    Influence Mobile

    Growth and mobile monetization across 150+ apps and games.

  5. 2014 to 2015

    Director, Head of Marketing

    CardTapp Inc.

    Company-wide marketing strategy for a mobile SaaS product.

  6. 2012 to 2014

    Program Manager

    Amazon

    Performance marketing programs and vendor management at scale.

  7. 2011 to 2012

    Sr. Business Product Manager

    GameHouse / RealNetworks

    Game monetization and go-to-market.

  8. 2011 to 2012

    Program Manager, Experimentation

    Expedia

    Ran the multivariate testing program across lines of business.

Hands-On Capabilities

Data & Modeling
XGBoost, classifiers, model evaluation on AUC, lift and calibration, feature engineering, real-time inference, SQL and BigQuery, Python, R
AI
LLM and agentic products, Vertex AI and Gemini, AI-driven campaign automation
Cloud & Engineering
Google Cloud across BigQuery, Cloud Run, Vertex, Firebase, global load balancing and Scheduler, Next.js and TypeScript, REST APIs, data pipelines
Marketing & GTM
Positioning and messaging, launch, pricing and monetization, sales enablement, competitive intelligence, lead scoring and routing, bidding marketplaces, CAC, LTV and ROAS, lifecycle and experimentation
Leadership
Team building and coaching, cross-functional influence, executive and investor communication

Education

B.A., Psychology, Western Washington University

About

Tue Pham

I started in experimentation, which is a useful place to start, because it teaches you early that most of your ideas are wrong and the only way to find out is to measure.

From there: performance marketing at Amazon, growth across a network of 150+ mobile apps, then insurance distribution at Fidelity Life, where I was the one buying leads at volume and getting judged on the cost per acquisition. That is the experience the rest of it is built on. I know what it feels like to spend real money on leads you cannot evaluate.

Two companies later I am still working on the same problem, now from the modeling side. I co-founded AveriSana, then RiskLync, where I am Chief Product Officer and also, most weeks, the person in BigQuery at eleven at night.

BurntOrange LLC is my own shop. It is how I take this work on for other companies without pulling them into RiskLync's roadmap. Based in Seattle, working remotely with clients anywhere in the US.

How I Work

01

I will tell you when the data does not support the claim

Most of this industry sells lift numbers without saying whether they came from a holdout, a retro, or a hopeful spreadsheet. I label mine. It costs me a little on the first call and saves everyone the argument later.

02

I build before I recommend

I still write the code, run the queries, and read the cloud bill. When I tell you something will work, it is usually because there is a version of it running somewhere I can show you.

03

Small number of clients at a time

Fractional work goes bad when it turns into four companies getting a quarter of someone. I keep the roster short enough that I am actually present in the meetings that matter.

04

The economics are the product

A model that is technically excellent and does not move CPA is a hobby. I start from the unit economics and work backward into what needs building.

Contact

Start a conversation

Tell me what you are trying to solve. If I am not the right person for it, I will say so, and I can usually point you at someone who is.

Based In

Seattle, WA. Remote engagements across the US.

What are you looking for?
Send Email

This opens your own mail app with the message filled in, so nothing gets submitted to a form you cannot see. Prefer to write it yourself? Use tue@burntorangellc.com.