Valid8 (Zenith Master): Deterministic Technical Claim Verification

Valid8 is an evidence-first hiring verification platform designed to validate technical claims with deterministic scoring and auditable proof trails. AI is used for extraction and interpretation, while final scoring remains rule-constrained.

Core Value

Reduces manual recruiter verification effort.
Produces explainable, evidence-backed candidate assessments.
Preserves deterministic behavior through explicit scoring formulas and traceable artifacts.

End-to-End Workflow

Ingest: Resume/profile/job-context inputs.
Audit: Validator engines analyze GitHub and profile evidence.
Enrich: AI extracts structured metadata from evidence.
Interpret: AI Judge evaluates claim-evidence consistency under constraints.
Aggregate: Deterministic weighted scoring and penalty logic.
Serve: Persisted results and evidence views for recruiter decisions.

Architecture

Backend: FastAPI, Pydantic, Python 3.13+, orchestrator-driven validator execution.
Frontend: React 19 + Vite + TypeScript recruiter dashboard.
Data Platform: Supabase Postgres + Storage with job-scoped persistence.
Security/Auth: Clerk identity model, schema validation, rate limiting, controlled key flow.

Scoring Model

Valid8 centers deterministic engine scoring (0-100 normalized), weighted aggregation, and explicit penalties. AI output is schema-validated and cannot bypass deterministic controls.

Final Score = (alpha x Engine Score) + (beta x Claim Score) - Penalties

AI Guardrails

Schema-first outputs only.
Unknown-over-guessing policy.
Prompt/version control to reduce drift.
AI assists interpretation; it does not replace deterministic scoring authority.

Canonical Positioning

Valid8 is a deterministic, evidence-first technical claim verification platform for recruiting workflows, combining multi-engine auditing, schema-constrained AI interpretation, and auditable verdict generation.