# SentryNode Gateway

> Semantic AI inference governance and FinOps platform. Reduces enterprise LLM spend by 77.6% through adaptive complexity-based model routing and hard budget guardrails. Intercepts agentic loops in under 40ms. Maintains 98.2% semantic quality parity. U.S. Patent Pending #63/982,542 (filed February 13, 2026). Apache 2.0 open core.

SentryNode Gateway sits between applications and model providers, scoring each request's semantic complexity and routing it to the most cost-effective capable model. It enforces hard budget caps, intercepts runaway agentic loops in real time, and provides a cryptographic audit trail (HMAC-SHA256) on every inference request. Built by Neha Dangwal, 9 years enterprise engineering at Cisco and Dell EMC.

## Product

- [SentryNode Gateway](https://sentrynodegateway.com): Full product overview — adaptive routing, circuit breakers, budget guardrails, voice governance, PII scrubbing
- [ROI Calculator](https://sentrynodegateway.com/roi-calculator/): Estimate cost savings from semantic routing for your inference volume
- [GitHub (open core)](https://github.com/nehadangwal/sentrynode-gateway): Apache 2.0 source, architecture, and deployment docs

## Technical Reference

- [README (https://github.com/nehadangwal/sentrynode-gateway/blob/main/README.md): Architecture overview, stack, deployment instructions, benchmark methodology

## Blog

- [The AI Tax Is Breaking Engineering Margins](https://medium.com/@nehadangwal/the-ai-tax-is-breaking-engineering-margins-a-17m-perspective-on-finops-7fce44983866): 77.6% cost reduction case study, $17M enterprise AI spend analysis, FinOps framework
- [The Sovereign Agent: Why Naked AI Execution Is an Enterprise Liability](https://medium.com/@nehadangwal/the-sovereign-agent-why-naked-ai-execution-is-an-enterprise-liability-f4f31a116493): Agentic loop governance, circuit breaker patterns, cryptographic audit trail rationale

## Optional

- [Design Partner Inquiry](mailto:partners@sentrynodegateway.com): Enterprise pilots and design partner program contact