Trivian Technologies builds AI governance middleware — the infrastructure layer between AI models and the production systems that deploy them.

Enterprises deploying AI agents face a real problem: models hallucinate, drift, and produce outputs that violate compliance requirements. No model provider solves this at the infrastructure level. We do.

Our flagship product, Syzygy Rosetta, is a containerised, API-first governance engine that evaluates every AI input and output before it reaches an end user — applying deterministic policy rules and ML-based risk scoring to return a structured, auditable decision in real time.

POST /evaluate

Every POST /evaluate call returns exactly these 8 fields:

{
  "decision": "allow | rewrite | escalate",
  "risk_score": 0.12,
  "confidence": 0.91,
  "violations": [],
  "rewrite": null,
  "reasoning": "Input within acceptable parameters for financial context.",
  "field_notes": [],
  "timestamp": "2026-03-21T14:32:00Z"
}

One endpoint. Every AI output evaluated, scored, and governed — before it reaches production.

What Rosetta is not: a chatbot, a foundation model, a UI layer, or a prompt wrapper.

What Rosetta is: infrastructure. It sits between your AI and your users and makes sure what goes through is safe, compliant, and auditable.

1 safety_layer.py Pre-classifies input — tags authority, manipulation, dependency, escalation patterns

2 policy engine Matches against industry-specific deterministic rules (finance / healthcare / general)

3 risk_scoring.py ML composite scorer — KeywordScorer + FeatureScorer + LLMScorer

4 Decision logic Threshold applied — allow / rewrite / escalate

5 Audit log Full evaluation appended to logs/evaluations.json

\# Clone the core engine

git clone https://github.com/Trivian-Technologies/syzygy-rosetta-originbase.git

cd syzygy-rosetta-originbase



\# Build and run (STILL IN PROD)

docker build -t rosetta .

docker run -p 8000:8000 rosetta



\# Evaluate your first input

curl -X POST http://localhost:8000/evaluate \\

-H "Content-Type: application/json" \\

-d '{

"input": "Your AI output here",

"context": {

"industry": "healthcare",

"user_id": "demo",

"environment": "staging"

}

}'

Enterprise teams deploying AI agents in regulated environments — healthcare, financial services, legal, compliance-heavy sectors — who need a governance layer they can audit, configure, and own.

Developers and AI engineers building on top of LLM providers who need structured output validation before their system acts on model responses.

Fintech companies requiring regulatory compliance enforcement

Healthcare platforms handling sensitive AI outputs

AI agent developers needing multi-agent governance

Regulated SaaS platforms with enterprise compliance requirements

Compliance and risk functions that need a documented, exportable audit trail of every AI decision made in production.

Syzygy Rosetta core engine is spec-compliant and confirmed working. The full 8-step evaluation pipeline is live — safety classification, deterministic policy enforcement, ML risk scoring, structured decision output, and persistent audit logging. Docker deployment and multi-agent sandbox testing are in progress.

Syzygy Rosetta is in active development. MVP delivery in progress.

For enterprise pilot enquiries, API trial access, or partnership discussions:

se@trivianinstitute.org