Solon delivers 99.9% Decision Accuracy in 0.7 Seconds Per Offense Across County, State, Federal, and National Criminal Databases, With Every Decision Fully Auditable and Reproducible

AUSTIN, Texas: #ArtificialIntelligence--SNH AI, the AI infrastructure company purpose-built for background screening, today announced the general availability of Solon, a domain-trained decisioning model designed to handle the full scope of criminal record review at production scale. Trained on nearly two million offense decisions and benchmarked against experienced human reviewers, Solon delivers actionable outcomes on 100% of offenses with a 99.9% accuracy rate, an achievement that was previously unreachable with existing technology.

Decisioning Speed and Accuracy

Solon’s 99.9% accuracy performance was produced from a training and evaluation corpus of 1.97mm offense records; each reviewed three times by subject matter experts before use. That review depth is the methodological foundation of model reliability. At just 0.7 seconds to receive a decision per offense, operations teams using Solon can instantly clear queues of thousands of records that previously required weeks of human review time.

Defensible Decisions, Reproducible Results

When facing an audit, a compliance team must be able to reproduce a decision made six months ago and support it with documented reasoning.

Every output from Solon includes the full reasoning that produced it: which identity signals were evaluated, which offense elements were weighted, which policy rule was applied, and how the determination was reached. That reasoning is stored, timestamped, and reproducible on demand. The same inputs will produce the same output with the same documented rationale, regardless of when the review occurs.

Solon is also fully configurable at the policy level. Each client's adjudication logic, including look-back periods, offense type exclusions, and position-specific rules, is encoded so that reportability determinations uniformly reflect the client's policy.

“Solon is engineered around real-world industry dynamics. It inherently understands the nuances of county criminal hits, the complexities of federal charges, and how client policies intersect with offense-level data. This deep operational expertise is exactly what drives our AI decision accuracy. Built for high-volume, high-stakes workflows, our system eliminates hallucinations to deliver decisions that are fully auditable, traceable, reproducible, and defensible." – Dr. Shams Syed, Chief AI Officer, SNH AI

Technical Specifications

Training corpus: 1.97mm criminal offense records, each reviewed three times by subject matter experts prior to use in model training and evaluation.

Record type coverage: County criminal, national criminal database, federal criminal, and statewide criminal databases.

Decision output: Each processed record produces a reportability determination, an identity match assessment, and a full reasoning chain with source citations at the record level.

Coverage: Actionable result on 100% of offenses processed.

Decision accuracy: 99.9%

Processing speed: 0.7 seconds per offense

Policy configuration: Fully configurable per client, including look-back periods, offense type exclusions, position-specific rules, and jurisdiction-specific logic.

Auditability: Every decision is reproducible. Reasoning and source citations are retained at the record level and accessible for compliance review.

Availability

Solon is available now for enterprise integration. For deployment information, API documentation, and pilot program access, contact us at info@snh-ai.com.

About SNH AI

SNH AI builds domain-specialized models and autonomous workforce systems. The company's models are trained on industry-specific operational data, purpose-built for production-scale deployment, and designed to meet the auditability, reproducibility, and compliance requirements of regulated workflows. SNH AI is headquartered in Austin, TX.

Fonte: Business Wire