Abe Burnett. OGDEN, UTAH · UPDATED AUG 2026

HEAD OF CENTER OF EXCELLENCE — IMMERSIVE  ·  OGDEN, UTAH

Abe Burnett

From raw data to defensible decisions.
Indices read by 82 million. Models that price a live insurance book. AI that survives audit.

70+
INTELLIGENCE ENGAGEMENTS DELIVERED IMMERSIVE · SEVEN MONTHS OF 2026
300+
ORGANIZATIONS BENCHMARKED IMMERSIVE · MITRE ATT&CK-ALIGNED
23B
DATA POINTS / MO — TECHNOLOGY INDEX PLURALSIGHT · 2016–23
11
RECOMMENDATIONS ON RECORD LINKEDIN
EXPERIENCE
FIG. R — EMPLOYMENT SPINE · 2016 → PRESENT ONE AXIS · YEARS ONLY HOVER OR TAB A ROLE — THE DOSSIER ANSWERS BELOW
PLURALSIGHT — SENIOR DATA SCIENTIST · 2016 — 2023 BURNETT TECHNICAL — FOUNDER / PRINCIPAL DATA SCIENTIST · 2023 — 2024 VANTAGE RISK — SENIOR AI DATA SCIENTIST · 2024 — 2025 IMMERSIVE — LEAD AI DATA SCIENTIST · 2025 — 2026 IMMERSIVE — HEAD OF CENTER OF EXCELLENCE · 2026 — PRESENT

SPAN WIDTH = TENURE, TO SCALE · BRIGHT SEGMENT = CURRENT ROLE, PROMOTED 2026 · DATES REAL, YEARS ONLY

2025 — PRESENT · CYBERSECURITY

Immersive

Head of Center of Excellence
2026 — PRESENT · PROMOTED

Founding leader of the company-wide specialist function — crisis simulation advisory, strategic intelligence, and API/platform integrations — built to turn one-off expert engagements into products the whole platform can reuse.

  • the specialist bench for the entire enterprise base — hundreds of organizations and the largest seven-figure national programs
  • stood up how the function runs in its first weeks: published service-level commitments, a service catalog for sales, and AI quality gates governing program completion
  • building the tooling layer — AI-assisted program generation, a crisis-simulation authoring platform with a reusable scenario library, and delivery automation that cuts the cost of every engagement after the first
Lead AI Data Scientist, Strategic Intelligence
2025 — 2026

Founded and ran the function: platform performance data turned into intelligence for some of the world's largest banks, insurers, consultancies, and government programs.

  • 70+ independent intelligence engagements personally delivered in seven months of 2026 — more than double the per-person average of the specialist team
  • AI-accelerated delivery: executive readiness reports, peer benchmarks, and threat-informed assessments for Fortune 500 security organizations in days instead of weeks
  • a threat-intelligence digest mapping live threat-actor activity to workforce readiness — validated by three marquee customers and selected for productization
  • a cyber-resilience benchmarking methodology aligned to MITRE ATT&CK across 300+ organizations, plus the anonymity governance that made cross-customer benchmarking possible
  • an ROI calculator and AI-assisted proposal system adopted by sales — proposals from days to ~30 min; recurring reporting automated, returning 60+ hours per cycle
2024 — 2025 · SPECIALTY INSURANCE

Vantage Risk

Senior AI Data Scientist
  • public-D&O actuarial pipeline modernized
  • systematic pricing bias surfaced behind premium leakage
  • GenAI claims classification — guardrails · human review · validation dashboards
  • multi-signal broker prioritization
  • company AI governance standards authored
2023 — 2024 · CROSS-SECTOR

Burnett Technical

Founder / Principal Data Scientist
  • independent consulting — nonprofits to Fortune 500
2016 — 2023 · EDTECH

Pluralsight

Senior Data Scientist
  • architected the Technology Index — 23B data points/mo · 850+ technologies ranked
  • syndicated coverage — 311 press pickups · 82M readers reached
  • led the index rewrite — refresh time −35%
  • led 5-engineer market-intelligence team — data sources +73% · data errors −40%
11 RECOMMENDATIONS ON LINKEDIN READ THEM →

The record above lists outcomes.
What follows is how the work gets done.

Selected workDemonstration · synthetic data

Ranking 850+ technologies

The three sections in this band are demonstrations — real methods, invented data, no client work. The technologies below are real; their trajectories are not.

The Technology Index reached more people than anything else I have built: a monthly ranking of 850-plus technologies, rebuilt from 23 billion data points a month, read by 82 million people, picked up by 311 newsrooms. I architected it at Pluralsight and led the five-engineer team that ran it — data sources up 73 percent, data errors down 40 percent, refresh time cut 35 percent.

More on this

The job was compression: 23 billion signals a month in, a short list of what rose and what fell out.

The demonstration below shows the shape — twelve technologies, twenty-four months, with the three biggest movers highlighted.

FIG. A — TECHNOLOGY RANK TRAJECTORIES 12 SERIES × 24 MO Demonstrationnot client work · synthetic data AS OF AUG 2026 HOVER THE FIGURETAP THE FIGURE — READOUT DOCKS HERE
RANK 1 IS THE TOP. EVERY LINE IS ONE TECHNOLOGY, RE-RANKED EVERY MONTH — SO A LINE THAT CLIMBS IS GAINING ON THE OTHERS, NOT MERELY GROWING. TAP A LINE FOR ITS CARD.
Selected work

Underwriting decisions

Every figure below is invented — this is not Vantage Risk work or data.

At Vantage Risk, I modernized the actuarial pipeline behind a live public-D&O book — the models that decide what a directors-and-officers policy should cost.

More on this

The rebuilt pipeline surfaced something the old one structurally could not: a systematic pricing bias — one direction, one segment band, the signature of a model error rather than noise. Premium had been leaking there for years.

I also ranked incoming broker submissions by expected value, so underwriters opened the right file first.

FIG. B — SUBMISSION TRIAGE + PRICING TRACE Demonstrationnot client work · synthetic data AS OF AUG 2026 HOVER A ROW OR MARKTAP A ROW OR MARK — READOUT DOCKS HERE
B-1 · BROKER SUBMISSIONS, RANKED BY EXPECTED VALUE — HOVER, TAP, OR TAB TO A ROW FOR THE RATIONALE
#SUBMISSIONPRIORITYEXP. VALUECONFIDENCE
EACH SEGMENT SHOWS WHAT THE MODEL EXPECTED TO LOSE AGAINST WHAT IT ACTUALLY LOST. RANDOM ERROR SCATTERS BOTH WAYS. FOUR NEIGHBORING SEGMENTS MISSING IN THE SAME DIRECTION IS A BROKEN MODEL, NOT BAD LUCK — AND THAT IS THE RED BRACKET.
1.00 MEANS THE PRICE COVERS THE EXPECTED LOSS. DARKER CELLS SIT FURTHER BELOW THAT LINE — THE POLICY IS SOLD FOR LESS THAN THE RISK IS WORTH.
Selected work

Evaluating an AI system

The model, the matrix, and the logs are invented — the sheet has the form of the real one; the counts do not.

Most companies say they use AI. Fewer can show whether it works. At Vantage Risk I shipped GenAI claims classification with guardrails at the boundary, human review where confidence is low, and validation dashboards executives could actually audit — then authored the company’s AI governance standards the system had to meet.

More on this

The artifacts that make it checkable are unglamorous: a model card, published next to the model — and a trip log, published next to the guardrails. Both are below.

FIG. C — QUARTERLY MODEL AUDIT · CLM-CLS v3.4 Demonstrationnot client work · synthetic data GUARDRAILS + HITL · AS OF AUG 2026 HOVER A CELL OR ROWTAP A CELL OR ROW — READOUT DOCKS HERE
CLASSPRF1ROUTE
2026-07-28 14:02Z · CONF 0.41 < 0.60 → ESCALATED TO REVIEWER
2026-07-28 09:47Z · PII PATTERN DETECTED → REDACTED · ESCALATED
2026-07-27 16:11Z · OUT-OF-SCOPE DOC CLASS → REFUSED · ROUTED MANUAL
2026-07-25 11:38Z · CONF 0.55 < 0.60 → ESCALATED TO REVIEWER
C-3 · MODEL CARD — CLM-CLS v3.4 Demonstrationnot client work · synthetic data FORM OF THE PRODUCTION CARD
Intended use
First-pass routing of incoming claim documents to six handling queues.
Out of scope
Coverage determination · reserving · any customer-facing output.
Decision rule
Auto-route at confidence ≥ 0.92; everything below routes to a human.
HITL rule
100% of low-confidence documents · 100% of the COMPLEX class regardless of confidence.
Guardrails
PII scrub · policy-bound prompts · refusal on out-of-domain input · full audit log.
Validation
Weekly drift check · quarterly audit (this sheet) · dashboards reviewed by claims ops.
HUMAN-IN-THE-LOOP: 100% OF LOW-CONFIDENCE CLASSES — BY STANDARD, NOT EXCEPTION. I WROTE THE STANDARD.

References

“Quickly became THE AI expert, not just for Data Science but for the entire company.”
Kimberly Venta AVP, Data Science Operations Lead · Vantage Risk Direct manager · LinkedIn
“Bridges technical and business — a true leader in his field.”
Ben Bowman Data Product Leader LinkedIn
11 recommendations on record. The two quoted here are verbatim. Read all 11 on LinkedIn →
Outside the day jobFive entries, stages dated

Built on my own time

I keep building after hours. Products, tools, and fiction — each entry below carries its stage, dated.

FlowState AI LIVE The discipline layer for AI-built software — shipped and selling. Trendline BUILDING Your body changes slowly. Your scale doesn’t. Guides live; iPhone app in build. InkForge BUILDING Writing tools for people who write for a living. Waitlist open. Heroes Wanted DRAFTING A fantasy-comedy serial about the gap between the job posting and the job. Vibe Coding for Fun and Profit DRAFTING How one person ships like a team. Full draft, in editorial review.
The build logs → The fiction →