Simple Machine Mind
SMSquared provides EvaluatorDPT, an enterprise AI decision system that delivers deterministic, policy-as-code decisions instead of probabilistic LLM outputs. The platform returns auditable YES/NO/TBD results with full reasoning traces, and is designed for safety-critical operations, compliance workflows, and AI agent guardrails. It achieved a 0.8252 Macro F1 across 44,597 held-out examples.
- Artificial Intelligence
- Regulatory & Compliance Technology
- Enterprise Software
- Software Only
Funding
Founders
Founder details are not available yet.
Product
Problem
Production systems that rely on probabilistic AI models (like LLMs) face unpredictable outputs, hallucination risks, and a lack of auditability, making it difficult to deploy AI in safety-critical or compliance-bound workflows. Teams need consistent, defendable decisions but lack a governed layer to enforce policies and explain outcomes.
Solution
EvaluatorDPT delivers a governed decision layer that executes deterministic, policy-as-code logic instead of probabilistic inference. The platform returns consistent YES/NO/TBD decisions with complete explainability metadata, including signals and reasoning traces, for every output. It enforces policy-aligned guardrails and supports intelligent escalation to human operators for edge cases where the system lacks confidence. Deployed as an inference-only API on Azure, it integrates cleanly with existing enterprise systems and maintains a secure posture with no data retention or storage of sensitive PHI/PII/PCI information.
Target Audience
Primary customers are enterprises running safety-critical operations (aviation, robotics, industrial systems), compliance-driven approval workflows (expenses, transactions), and teams needing guardrails for LLM and AI agent outputs in production.
Features
- Deterministic decision engine: same input produces the same decision every run, eliminating prompt drift and hallucinations
- Policy-as-code guardrails: decisions are shaped and constrained by programmable policies for compliance and safety
- Full audit trail: every decision logs signals, vectors, and reasoning traces for transparency and regulatory defense
- TBD escalation: uncertain edge cases automatically route to human review rather than forcing a binary outcome
- Production-ready REST API: plug-and-play, inference-only integration on Azure with enterprise security
- No data retention: no PHI/PII/PCI storage and no open weights, ensuring data sovereignty and privacy
- Measurable performance: achieved 0.8252 Macro F1 across 44,597 held-out examples