Frank Speiser
Co-Founder, Managing Partner
Product, platform, applied ML
- Co-Founder & CPO, SocialFlow (acquired)
- Builder of agentic-AI systems for high-volume decisioning
- Two decades shipping data-intensive software at scale
Agentic-native consulting
Government systems that decide faster, pay correctly, and explain themselves.
Vardr Partners modernizes state and federal benefits programs with agentic AI — anti-fraud detection, neural-network signal discovery, and decisioning copilots — shipped behind feature flags, never a flag-day cutover.
- Avg. eligibility-decision reduction
- 60–90%
- Cutover style
- Feature-flagged
- Decisioning audit trail
- Per-record
- Section 508 conformance
- WCAG 2.1 AA
Built for the programs that move money
Research & reference
The arithmetic behind the decisions states are making now.
Reference data and decision frameworks we build for our own engagements, published because the statutes apply to every state at once. The data is on the page; the formatted document is a short email confirmation away.
SNAP integrity
The Six Percent Decision
The five paths a state can take on the SNAP cost-share liability, each with what it costs, how long it takes to bite, and what you would have to believe for it to be the right call — plus the sixteen jurisdictions where one point of error rate is worth a full tier.
9-page PDF · Sep 2026
Read the data and get the PDF
SNAP integrity
SNAP FY2028 Cost-Share Data Briefs — One Page per Jurisdiction
A one-page FY2028 cost-share brief for each of the 53 reporting jurisdictions: enrollment, annualized issuance, FY2025 payment error rate, the statutory tier that rate produces, the projected annual state share, and what crossing the next threshold is worth. Computed from the same reconciled dataset as the national analysis — no narrative, no per-state prose.
53 jurisdictions · Sep 2026
Pick a jurisdiction and get its one-pager
SNAP integrity
SNAP OBBBA §10105 Cost-Share Obligation by State — FY2028 Projection
What OBBBA §10105 cost-sharing does to a state budget, and the projected annual obligation for all 53 reporting jurisdictions — ranked by FY2025 payment error rate, with the statutory tier mechanics and every caveat that belongs on the figures.
53 jurisdictions · 5-page PDF · Sep 2026
Read the data and get the PDF
Email verification required; marketing emails optional.
01Market scale
The markets we work in are enormous and growing.
Four addressable segments. Combined TAM north of $5T in annual program dollars with a weighted SAM CAGR above 7%. The systems Vardr operates on are not edge cases — they are the load-bearing decision engines of American government.
Total Addressable Market
$5.6T
Program dollars in scope
Serviceable Addressable Market
$35.0B
Annual modernization + advisory spend
Weighted CAGR (5-yr SAM)
8.1%
Across all four segments
Market segments — annual scale
TAM (program dollars) vs SAM (services spend) vs SOM (3-yr Vardr-obtainable). Log scale.
TAMSAMSOM
State Medicaid
CAGR 6.2%
TAM
$870.0B
SAM
$6.5B
SOM
$180M
Federal Civilian
CAGR 8.5%
TAM
$3.4T
SAM
$22.0B
SOM
$320M
State Tax & Revenue
CAGR 7.2%
TAM
$1.1T
SAM
$2.4B
SOM
$65M
Program Integrity & OIG
CAGR 9.4%
TAM
$233.0B
SAM
$4.1B
SOM
$95M
5-year SAM growth projection
Compounded at each segment's rolling-5yr CAGR. Hover a segment to highlight.
Sources: CMS NHE, OMB historicals, GAO improper-payment series, FTA state collections. Order-of-magnitude estimates; full methodology on hover per segment.
02Services
Five practices. One operating model.
Each practice ships into agency environments with the same discipline: versioned models, reproducible decisions, traffic-shadowed migration, and human-in-the-loop review where it matters.
- 01
Benefits Verification & System Modernization
Re-platform legacy eligibility and verification systems without service disruption.
Capabilities - 02
Audit Simulation & Integrity Forecasting
See your OIG findings before the auditors do.
Capabilities - 03
Anti-Fraud & Fraud Detection
Stop improper payments before they go out the door.
Capabilities - 04
Decisioning Assistance
Augment caseworkers without replacing their judgment.
Capabilities - 05
AI Guidance & Implementation
From OMB memo to production system, in a defensible plan.
Capabilities
02Open Workbench
Original methodology, published in the open.
We publish the frameworks we use on engagements. Procurement teams can review them before a single email is exchanged. Other firms keep this in a sales deck. We don't.
VARDR / METHODOLOGY
Agentic Maturity Model
- L0
Rules─ tier 0
Hand-written rule engines, static dashboards
- L1
Augmented─ tier 1
ML scoring layered onto rules; manual review
- L2
Assisted─ tier 2
Decision copilots; citations and audit trails
- L3
Agentic─ tier 3
Multi-step agents under human policy constraints
- L4
Autonomous─ tier 4
Self-monitoring agents; closed-loop remediation
Most agency systems live here ↑ between L0 and L1.vardrpartners.com/approach
03Leadership
Principal-led. One accountable signature on every engagement.
Vardr is run by the people you'll meet on day one. Managing partners Frank and Payton sign every statement of work and remain involved through delivery. Partner Lewis Gossett leads state affairs and regulatory strategy. Not a partner-bait-and-junior-switch.
Payton Jonson
Co-Founder, Managing Partner
Government delivery, program integrity
- Hands-on operator across state benefits modernization
- Procurement, compliance, and accountable-cutover discipline
- Bridges policy, program, and engineering teams
04Posture
How we operate
- We never run on a flag-day cutover. Every change ships behind a feature flag with parallel-run telemetry.
- Every model and decision we ship is reproducible from versioned data, code, and prompts.
- We will not propose a solution we cannot also operate. If we cannot run it, we do not recommend it.
- We hold continuity of service as a hard constraint above velocity.
05Insights
Our current thinking, not last quarter's whitepaper.
When a program can't afford another failed modernization, call us first.
45-minute principal-level briefing. NDA on file before discovery.