Platform

Every layer of AI governance, engineered as one system.

AIOPS is not a dashboard bolted onto spreadsheets. It is a distributed operating system — API, durable workflow engine, streaming monitors and policy plane — purpose-built to govern machine learning in production.

Architecture

Four planes. Zero gaps.

Each concern is isolated into its own plane with explicit contracts between them — so a failure in one can never silently corrupt another.

Data plane

PostgreSQL with row-level security and pgvector, Redis for revocation and real-time streams, Qdrant for semantic retrieval, and WORM object storage for immutable evidence retention.

PostgreSQL + RLS · Redis · Qdrant · WORM object store

Policy plane

160+ compliance rules across ISO 42001 and NIST AI RMF, enforced by Open Policy Agent. Policies are authored visually, compiled to Rego and published through git pull requests with full review history.

OPA · Rego · visual policy builder · git publishing

Evidence plane

Assessment results, fairness metrics and SHAP explanations are packaged into Ed25519-signed archives, chained into an append-only SHA-256 ledger, with optional blockchain anchoring for external timestamping.

Ed25519 signing · SHA-256 ledger · anchoring

Security plane

Fail-closed PII scanning on every ingest, semantic prompt-injection detection, streaming output redaction, and SPIRE-issued workload identities enforcing mutual TLS between every service.

Presidio fail-closed · injection detection · SPIRE mTLS
Modules

What ships in the box.

Model inventory

Governed lifecycle state machine for every model, from registration to retirement, with immutable transition history.

Assessment engine

Deterministic ISO/NIST scoring, EU AI Act classification, fairness metrics and exact SHAP explainability per assessment.

Evidence builder

One-click signed evidence ZIPs formatted for auditors and insurance underwriters alike.

Real-time monitors

Streaming agents for PSI drift, demographic parity, equalized odds and latency — pushed to dashboards over live server-sent events.

LLM gateway

A governed proxy for LLM traffic with guardrail scanning, ensemble routing, spend tracking and per-request policy checks.

Policy builder

No-code policy composition that compiles to OPA Rego and publishes to your git repository as reviewable pull requests.

Shadow AI discovery

Scanners that detect unregistered AI usage across cloud platforms and surface it for onboarding or shutdown.

Compliance copilot

A retrieval-augmented assistant grounded in your actual policies, assessments and framework corpus — advisory, never authoritative.

Reporting

Board-ready PDF and DOCX reports, model cards and posture scoring generated straight from live governance data.

Integration

Meets your stack where it runs.

Instrument models with the open-source Kafka SDK, connect cloud AI platforms through managed connectors, or drive everything over the REST API.

  • Native connectors for AWS, Azure and GCP AI platforms
  • Open-source Kafka SDK streams model events into governance
  • REST API with JWT auth for every platform capability
  • Webhooks for assessments, alerts and lifecycle transitions
  • OpenTelemetry traces and Prometheus metrics out of the box
# Stream model events into AIOPS governance
from aiops_kafka_sdk import ModelEventProducer

producer = ModelEventProducer(
    model_id="credit-risk-v4",
    tenant="acme-bank",
)
producer.emit_prediction(
    features=payload,
    outcome=decision,
# drift & bias monitored in real time

Walk the architecture with an engineer.

Our demos are run by the people who built the platform — bring your hardest deployment questions.