VAST Data unveiled PolicyEngine and TuningEngine, two new services designed to help the next-generation VAST AI OS at VAST Forward 2026, aiming to meet the needs of organizations scaling mission-critical AI.
PolicyEngine and TuningEngine by VAST Data work together to build AI systems that are trusted, explainable, and continuously learning. PolicyEngine governs agentic activity, while TuningEngine manages model tuning, powering automatic learning loops aligned with organizational goals.
As AI workflows increasingly access organizational data to generate responses, communicate between agents, and log events, the risk of data spillage rises without fine-grained controls. VAST PolicyEngine addresses this with an inline policy enforcement engine that governs agents’ access to memory, tools, knowledge bases, and other agents using explicit permissions and AI-driven context. Enforcement occurs before actions execute, while tamper-proof logs ensure all decisions remain observable, explainable, and auditable.
VAST TuningEngine leverages the serverless AgentEngine runtime to coordinate multi-agent workflows, model invocation, and tool usage. It captures outcomes from agentic pipelines, applies curated feedback, and supports LoRA fine-tuning, supervised tuning, and reinforcement learning.
New candidate models are evaluated, benchmarked, and deployed, kickstarting learning loops that continuously improve system performance. TuningEngine also integrates with NVIDIA’s NeMo Data Designer to fine-tune Nemotron open models.
Jeff Denworth, Co-Founder at VAST Data, said that, “Just as people are always learning, so should tomorrow’s applications. With the introduction of PolicyEngine and TuningEngine, the VAST AI Operating System has become a thinking machine that customers can deploy wherever they compute, a machine that safeguards every interaction and learns from every outcome, bringing the power of AI within reach of every organization.”
These capabilities mark a major step toward VAST Data’s vision of a ‘thinking machine’ that evolves through interaction with real-world data. Since 2016, VAST Data has been building toward this system, and with PolicyEngine and TuningEngine, the AI OS now completes a closed loop of observing, reasoning, acting, evaluating, and improving, all while maintaining security and explainability. Both engines are set for release by the end of 2026.
