Our AI Foundry
Duforest doesn’t just build software; we engineer AI systems with enforced data residency, controlled execution, and verifiable behaviour.
Agentic Engineering
Our agentic engineering framework provides the structural foundation for building autonomous systems that can reason, plan and act within tightly regulated financial and life-science environments. It combines formalised regulatory context, controlled multi-agent execution, and forward-looking compliance intelligence to ensure every agent operates within defined legal, safety and governance boundaries. This approach allows complex tasks to be decomposed, orchestrated and executed with full auditability, jurisdictional assurance and operational integrity.
Compass
Compass translates complex regulatory requirements—whether financial conduct rules, clinical governance standards, GxP controls, or data-protection obligations—into a formal constraint architecture for agent behaviour. Domain-specific regulations are decomposed into machine-readable policies governing data residency, permissible actions, evidence requirements, and approval workflows. This yields a deterministic operating envelope in which agents must function, ensuring every autonomous step adheres to sector-specific compliance, safety and traceability standards.
Conductor
Conductor provides the execution control layer for assembling agentic systems across regulated environments. It manages the sequencing, dependency resolution and inter-agent communication of specialised reasoning, retrieval, validation and action agents. Through policy-bound tool-calling, sandboxed execution paths, and immutable audit trails, Conductor guarantees that complex workflows—whether risk assessments, KYC analyses, pharmacovigilance reviews or clinical documentation tasks—remain explainable, reproducible and aligned with the organisation’s operational controls.
Catalyst
Catalyst extends agent capability with continuous regulatory foresight tailored to both financial and life-science domains. It performs global horizon scanning across legislative pipelines, supervisory guidance, medical-device directives, clinical standards and emerging compliance frameworks. Using scenario modelling and long-range risk analysis, Catalyst anticipates regulatory shifts and encodes future-state requirements into the architecture. This ensures that agent workflows remain resilient to forthcoming changes—supporting proactive adaptation, audit readiness and strategic compliance planning.
Sovereignty First
Our Foundry infrastructure is built on Google Cloud to enforce strict data residency and sovereignty for regulated financial and life-science organisations. Workloads are deployed into region-locked environments so model execution, storage and agentic workflows stay inside the jurisdictions you specify, while still taking advantage of Vertex AI and the wider Google Cloud platform.
Sovereign Cloud Execution on Google Cloud
We deploy models and agents using Vertex AI and Vertex AI Agent Builder, including Vertex AI Agent Engine for managed agent runtimes, with all resources pinned to specific regions. Network boundaries are enforced with VPC Service Controls and Private Service Connect, ensuring that traffic to Google-managed services remains inside defined perimeters and cannot be exfiltrated outside your authorised projects or regions.
Air-Gapped and Restricted-Connectivity Deployment
For high-assurance workloads — from trading and risk engines to clinical pipelines, pharmacovigilance and genomics — we support tightly restricted or offline deployment patterns. Agents built with Google’s Agent Development Kit (ADK) can be containerised and run on Cloud Run or Vertex AI Agent Engine with private ingress only, zero public endpoints and no default egress. For environments requiring stronger isolation, we combine private VPC-only architectures with Confidential VM-based execution to protect data in use as well as at rest and in transit.
Data Residency, Encryption & Locality Assurance
Data locality is enforced cryptographically using Customer-Managed Encryption Keys (CMEK) in Cloud KMS, so you retain ownership of the keys that protect your datasets, model artefacts and agent traces. Combined with regional storage, per-service access controls and audit-ready logging, this gives verifiable evidence that sensitive financial or clinical data remains within the jurisdictions and regulatory regimes you select, and that every access can be traced for inspection or assurance
Technology Partners
Our R&D stack is built on proven, production-grade platforms from Google Cloud and OpenAI, enabling us to engineer agentic systems that meet the security, governance and performance requirements of regulated financial and life-science environments.
Google Cloud
We deploy agentic workloads using the Agent Development Kit (ADK), Vertex AI, and regionalised VPC-secured architectures. This provides controlled execution, model-lifecycle governance, and strict adherence to data-residency requirements.
OpenAI
Where appropriate, we integrate OpenAI frontier models for advanced reasoning, long-context analysis and specialised domain tasks. These models are incorporated under the same governance, auditability and locality controls defined by our Foundry framework.
Hybrid Model Strategy
Our platform is model-agnostic. For each agentic workflow we select — and benchmark — the most suitable model or model family, whether from Vertex AI, OpenAI, or curated open-source systems. This ensures optimal accuracy, performance and compliance while maintaining full traceability of model behaviour and data flows.
Ready to Deploy?
Explore our portfolio of sovereign AI ventures or contact us to discuss a strategic partnership.