Chapter 01 — Our Mission

Built for the youth
every other platform
was built without.

To eliminate mental health disparities by building AI that understands how diverse youth actually communicate about distress, trauma, and healing.

We are building the first behavioral health platform designed by and for BIPOC, LGBTQ+, and first-generation youth communities. Not retrofitted. Not translated. Built from inside the language, the culture, and the lived experience — from the beginning.

Our story Partner with us
42+
Culturally specific distress signals detected across five behavioral dimensions
CulturalBERT-VLAP signal taxonomy
2,400+
AAVE and youth vernacular tokens in the CulturalBERT-VLAP model vocabulary
Community-sourced, clinically annotated
94%
Sensitivity on high-distress signal detection — clinical validation in progress
IRB study active · University of Maryland
$0
Cost to members in all partner program deployments
Org and payer funded · Access is not a billing question
Chapter 02 — Our Story

A realization that
became a refusal.

Vasl Health was built from a specific observation: existing mental health AI doesn't understand how BIPOC, LGBTQ+, and first-generation youth actually communicate about distress. Generic models are trained on majority-White internet text. They consistently miss AAVE, code-switching, youth vernacular, and the coded language communities develop when the dominant culture's vocabulary isn't safe, available, or accurate. The gap isn't fine-tuning. It's architecture. We built something different.

2023
The Founding Insight

Our founders discovered that mainstream AI models consistently missed culturally coded distress signals — producing dangerous gaps in clinical awareness for diverse youth populations. The problem wasn't access to AI. It was that the AI available had never been trained on the communities it was being deployed to serve. That realization became the mandate: build the language model that should have existed already.

The Gap Identified
2024
VLAP Development

We began developing CulturalBERT-VLAP — the Vasl Language Analysis Platform — training a BERT-architecture model on 198,000+ culturally specific mental health language samples drawn from BIPOC and LGBTQ+ youth communities. Every training sample was clinically annotated by licensed clinicians with community competency training. The 2,400+ token vocabulary extension was built from community language — not from clinical corpora or majority-population datasets.

CulturalBERT-VLAP V1
2025
Pilot Success

Pilot programs across community health centers, school-based programs, and university partnerships demonstrated 79.5% 30-day retention — nearly three times the industry average — and 42% improvement in PHQ-8 depression scores at 90 days. The data proved what the framework promised: when the platform understands how youth actually communicate, they actually stay. An IRB study with the University of Maryland began validating clinical signal accuracy in production deployment.

79.5% Retention · 42% PHQ-8 Improvement
2026
National Scale

Expanding partnerships with healthcare systems, school districts, and community organizations to bring the Vasl platform and VLAP clinical interpretation infrastructure to communities nationwide. Mayo Clinic Platform_Accelerate program participant. IRB study ongoing. Behavioral health data infrastructure being developed with Mayo Clinic Platform to support VLAP model development, validation, and LEP population baseline work across a two-year arc.

Mayo Clinic Platform_Accelerate · National Expansion
Chapter 03 — How VLAP Works

The language youth
actually use.
Finally read correctly.

CulturalBERT-VLAP — the Vasl Language Analysis Platform — detects culturally specific distress signals across five behavioral dimensions. It processes language youth share through their care channels and surfaces interpretive context to licensed clinicians before sessions begin. It does not diagnose. It does not prescribe. It does not respond to members. It is a precision interpretation instrument — clinician-facing only.

Signal context surfaces to licensed clinicians only — never to members, school staff, or administrators

Pattern — AAVE Indirect Hopelessness
HOP-03CCM-09
What the member said
"idk why i even try anymore tbh, can't keep doing this no more fr."
What VLAP detects — surfaced to clinician
Indirect hopelessness expressed through AAVE construction. "No more fr" is an authenticity escalator — not informal emphasis. Standard NLP classifies this as low-confidence text and generates no signal. VLAP reads: sustained distress beneath a vernacular frame. HOP-03 + CCM-09 flagged. Clinician receives dimensional context before next session.
Non-diagnostic · Clinician-facing only · Human review initiates all clinical response
Pattern — Pre-Disclosure Minimization
CCM-04ISO-04
What the member said
"it's not that deep but lowkey been struggling since school started."
What VLAP detects — surfaced to clinician
CCM-04: pre-disclosure minimization pattern. "It's not that deep but" is a documented pre-disclosure frame in youth language — a hedge before authentic disclosure. Standard models reduce signal weight at the minimization phrase. VLAP increases it: the hedge is the signal. Clinician arrives prepared to open this specific thread.
Non-diagnostic · Clinician-facing only · Human review initiates all clinical response
Pattern — Coded Self-Harm Language
SHA-03CRS-02
What the member said
"been thinking about unaliving lately ngl."
What VLAP detects — surfaced to clinician
"Unaliving" is a coded community term for self-harm ideation — developed to avoid content filters. It does not exist in standard NLP training data. It is in the VLAP vocabulary. SHA-03 + CRS-02 detected. Signal surfaced for immediate clinical supervisor review. A licensed human clinician determines the response — not the platform.
Critical priority · Surfaced for immediate clinical supervisor review · Human determination of response · 90-minute human SLA
198K+
Culturally-specific
training samples
5
Behavioral signal
dimensions analyzed
0
Diagnostic outputs.
Ever.
Chapter 04 — Trusted Partners

Organizations
that share
the commitment.

Vasl works with healthcare systems, research institutions, and community organizations that are already doing the work of mental health equity — and need the platform infrastructure to do it better. These are not logo partnerships. They are active deployments, research collaborations, and clinical validation relationships.

Healthcare & Research
University of Maryland
Active IRB study validating VLAP clinical signal accuracy in production deployment
University of Baltimore
Clinical validation and outcomes research partnership
Mayo Clinic Platform_Accelerate
Behavioral health data infrastructure for VLAP model development and LEP population baseline work
Community Organizations
Thread YMCA
Youth mentorship and behavioral health support programs
Boys & Girls Clubs
Youth development and mental health support integration
Enoch Pratt Library
Community access and digital equity programming
Chapter 05 — Validation & Compliance

The standards
the work
requires.

Clinical platforms serving minors, Medicaid populations, and school-based programs operate in one of the most demanding compliance environments in software. Vasl is built for that environment — not retrofitted for it. Every standard below is structural, not policy-based.

HIPAA
Full technical safeguard implementation. BAA required for all partners.

Complete HIPAA technical safeguards across all platform components. Business Associate Agreement required before any partner deployment. Annual SOC 2 Type II audit. VLAP processes language in-memory without verbatim storage. All staff with PHI access complete annual HIPAA training.

PHI protection protocols Audit logging and access controls Staff HIPAA training — annual
FERPA
Student education records protected. No individual data shared with school staff.

Vasl operates as a direct service provider to students — not as an agent of the school district. Student health records are protected under HIPAA architecture and never disclosed to administrators, teachers, or parents without explicit student consent. FERPA-compliant data handling for all school and university partnerships.

Educational record security Parental consent processes — minor-specific No individual data to school administrators
SOC 2 Type II
Annual third-party security audit. Full report available under NDA.

SOC 2 Type II certification covering security, availability, and confidentiality trust service criteria. Independent third-party audit conducted annually. Report available to clinical and institutional partners under NDA upon request. Processing integrity and confidentiality measures fully documented.

Security and availability controls Processing integrity Confidentiality measures
WCAG 2.1 AA
Full accessibility compliance across all platform portals.

The Vasl platform meets WCAG 2.1 Level AA accessibility standards across all portals — Member App, Coach Portal, and Client Org Portal. Screen reader compatibility, keyboard navigation, sufficient color contrast, and alternative text are implemented and tested. Accessibility audits conducted with each major platform release.

Screen reader compatibility Full keyboard navigation Color contrast standards — all portals
Chapter 06 — What We Value

What guides
every decision
we make.

Not principles on a wall. The standards by which we evaluate every product decision, partnership, and architectural choice — applied before we build, not explained after.

01
Cultural Authenticity

We build with and for the communities we serve — not for a generalized version of them. Authentic representation isn't a design requirement. It's the reason the platform works differently from everything built before it.

02
Evidence-Based

Every feature is backed by rigorous research, clinical validation, and real-world outcomes data. Claims are stated with methodology. IRB study data, not projected benchmarks. The 42% and 79.5% figures are from deployed cohorts, measured against pre-enrollment baselines.

03
Community-Centered

Youth voices guide our development process. Not as a consultation checkbox — as a structural requirement. The peer group norms, the VLAP training corpus, the care model sequencing: all shaped in active partnership with the communities the platform serves.

04
Privacy First

We protect user data with the highest security standards — and maintain complete transparency about what the platform does and doesn't do with it. VLAP processes language in-memory without verbatim storage. Individual member data never reaches school staff or org administrators. The architecture enforces this. Policy does not.

05
Health Equity

Written into the Delaware Public Benefit Corporation charter — not the marketing copy. The commitment to advancing health equity for communities of color, LGBTQ+ individuals, and low-income and underserved families is a legal obligation that survives every growth stage and investor conversation.

06
Human in the Loop

AI informs. Humans decide. Always. No automated clinical action follows from VLAP output. No alert fires without a human reviewing it. The platform is designed so that clinical judgment is never delegated to an algorithm — and this boundary is architectural, not a policy preference.

Chapter 07 — The Impact So Far

What the data
says — and what
comes next.

The numbers below are from active pilot deployments — not projected, not benchmarked from comparable platforms. Measured from Vasl's own member population. The work is ongoing. The IRB study is active. The scale is beginning. If your organization is ready to be part of what comes next, we want to hear from you.

Partner With Us View Careers
79.5%
30-Day Retention
vs. 40–50% industry average
42%
PHQ-8 Improvement
90-day cohort measurement
94%
Signal Sensitivity
High-distress detection accuracy
3wks
Median to First Support
vs. 9 months nationally