Founder and Systems Architect at OrvixLabs

I build private AI systems that investigate, challenge their own conclusions, preserve the evidence behind every claim, and state clearly what they cannot confirm.

• Evidence before confidence — a conclusion cannot be stronger than the material supporting it.
• Adversarial verification — results are challenged independently before reaching the user.
• Missing data stays missing — unknown information is declared, never silently invented.
• Traceability by design — claims preserve their sources, dates, confidence and unresolved objections.
• Structural coherence — a final report cannot contradict its own evidence or audit.
• Vendor independence — a system should remain explainable and sustainable without permanent dependence on its creator.

These are not interface features. They are architectural constraints.

A domain-agnostic deliberative reasoning engine.

Independent agents produce, verify, challenge and consolidate conclusions. Credibility is limited by available evidence, and a coherence lock prevents unsupported or contradictory conclusions from being presented as reliable.

→ View the public technical documentation

A divergent discovery engine designed to generate competing hypotheses, alternative explanations and unexpected investigative paths.

CASANDRA explores possibilities. It does not decide which possibilities are true.

A deterministic verification engine built around an immutable claim ledger, explicit commands and events, reproducible execution and auditable state transitions.

Its purpose is not to discover possibilities. Its purpose is to determine which claims survive verification.

CASANDRA explores. CERBERUS verifies.

I use these architectural principles to build:

• Commercial intelligence platforms
• Company and opportunity investigation systems
• Audited decision-support systems
• Messaging, email, voice and SMS integrations
• Private and self-hosted AI infrastructure
• Multi-instance products with isolated deployments

I work across the complete system:

• Reasoning and orchestration engines
• Data contracts and audit models
• Infrastructure and deployment
• Security and tenant isolation
• Telephony and messaging integrations
• Operational interfaces
• Monitoring and traceability

I build systems that handle real conversations, business decisions, money and regulatory constraints, where invented information is not merely an error. It is a liability.

Private AI systems, reasoning engines, commercial intelligence products and technical architecture for organizations that need AI to be auditable, explainable and operationally sustainable.

Based in Argentina. Working remotely worldwide.

• OrvixLabs — orvixlabs.com
• Personal website — carlosperasso.ar
• IRIS SCE — iris-sce.com