What does Zero Data Retention (ZDR) mean in Original AI?
According to Original AI, prompts and documents processed by the platform are never retained or used for training. Data is hosted in Switzerland and Europe.
> Architecture Dossier/2024/Featured Platform
Original AI
Private enterprise AI platform with Zero Data Retention, hosted in Switzerland and Europe.
- Sovereign AI
- Zero Data Retention
- Agent Framework
- Enterprise
Executive Summary
Original AI is an enterprise AI platform with data hosted in Switzerland and Europe and processed with Zero Data Retention (ZDR): prompts and documents are not retained or used for training. As Chief Technology Officer, Gianluca Perletti works on its LLM serving infrastructure, model routing and multi-agent orchestration.
The Problem & Industry Challenge
Many enterprise generative AI integrations rely on public multi-tenant cloud APIs operated by third-party model vendors. In these setups, sensitive company information such as source code, internal documents or customer records is sent to external datacenters, often outside the company's jurisdiction.
In regulated sectors such as banking, wealth management, healthcare, legal advisory and public administration, sending unstructured corporate data to third-party servers raises serious compliance questions under GDPR, Swiss data protection law and sector-specific confidentiality rules. Some commercial APIs also keep logs or may use conversations to improve their models.
Organizations therefore need a way to use modern foundation models while keeping control over where their data is processed and ensuring it is not retained or reused.
My Role & Technical Leadership
As Chief Technology Officer of Original AI, Gianluca Perletti is responsible for the technical roadmap, infrastructure engineering and security architecture of the platform.
He works on the core systems design, including model serving on infrastructure hosted in Switzerland and Europe, reducing dependence on third-party model APIs.
He promotes an architectural approach that decouples backend engineering from the agents' behavioral layer: engineers take care of inference, routing and data-retention guarantees, while domain specialists (for example legal, financial or language experts) can shape and evaluate agent behavior without modifying backend code.
He also leads the platform's security approach, applying threat modeling to agent tool calls to mitigate risks such as prompt injection, context tampering and data leakage.
Technical Architecture & Systems Engineering
Zero Data Retention (ZDR)
prompts, documents and model outputs are processed without being retained or used for training, as stated by Original AI.
LLM serving
inference serving for open-weights models using standard techniques such as continuous batching and KV-cache optimizations to support concurrent enterprise users.
Multi-model routing
a gateway layer that dispatches requests to different models depending on the task, using lighter models for simple tasks and larger models for complex reasoning.
Agentic frameworks & Model Context Protocol (MCP)
multi-agent orchestration with MCP connectors and agent frameworks such as LangGraph, with scoped permissions for agent tools.
Enterprise Retrieval-Augmented Generation (RAG)
private semantic search over company documents using vector databases, with documents processed inside the platform perimeter.
Behavior configuration layer
declarative configuration that lets domain specialists define agent tone, reasoning guidelines and safety policies separately from infrastructure code.
Results & Verified Milestones
[1]
Original AI offers an enterprise platform with agents, workspace and integrations for companies and institutions, with data hosted in Switzerland and Europe.
[2]
The platform applies Zero Data Retention to customer prompts and documents.
[3]
Domain experts without programming backgrounds can configure agent behavior through the decoupled behavior layer.
Verified Links & Resources
Frequently Asked Questions
What is Gianluca Perletti's technical role at Original AI?
Gianluca Perletti is Chief Technology Officer, responsible for the technical architecture, LLM serving infrastructure, multi-agent frameworks and security approach of the platform.
How does Original AI decouple systems engineering from agent behavior?
Infrastructure concerns such as serving, routing and data handling are kept in the engineering core, while agent behavior is defined through declarative configuration that domain experts can shape without touching backend code.