Platform for asset managers and asset owners
The ZTA platform provides the governed AI foundation beneath Investment Research and other institutional investment workflows. It is designed for asset managers and asset owners that need private inference, entitlement-aware data access, multi-model governance, Human Judgment, and an institutional learning record.
What ZTA delivers
The ZTA deployable reference architecture is built on open-source software and open-weight models, designed to run on infrastructure the institution controls. Components can be replaced over time without surrendering the institution’s data, governance, evaluation standards, decision history, or accumulated intelligence.
Open-source first. Vendor-flexible by design. Institution-owned by default.
A governed AI operating environment
ZTA separates the institution’s data, governance, workflows, and accumulated intelligence from any single model provider. Models can change as better ones emerge without requiring the institution to rebuild its AI operating environment.
Sensitive institutional data and research remain inside the institution’s chosen control boundary.
Models, evidence, policies, Human Judgment, and decisions remain traceable.
AI usage does not have to create perpetual third-party token dependency.
Better models can be introduced without redesigning the surrounding institutional architecture.
Institution-controlled architecture
Data, models, compute, agents, governance, and Human Judgment should not operate as disconnected technology layers. ZTA brings them into one institution-controlled architecture designed to share capability without creating gaps in control.
Shared structure.
Replaceable components.
Institution-owned intelligence.
Maximum capability. Minimum structural waste. No gaps in control.
ZTA AI Control Plane
The control plane governs how AI operates inside Investment Research and other institutional investment workflows: what intelligence may access, which models and agents may act, how much autonomy is permitted, and when Human Judgment becomes mandatory.
Applications · Identity · Approved Data
Read-only access, approved evidence, provenance, and institutional boundaries.
Human and agent identity, task-scoped authority, and tool permissions.
Baseline risk, runtime escalation, autonomy boundaries, and mandatory controls.
Approved model pools, benchmark-led routing, and cost-performance policy.
Deterministic checks, Model Council when required, and Human Judgment.
Open-source software · Approved open-weight models · Inference · Knowledge · Logging
Institution-selected infrastructure posture on NVIDIA infrastructure.
Open by design
No individual model, software component, infrastructure provider, or operating partner should become the institution’s permanent dependency. ZTA separates replaceable technology from the durable institutional assets that need to survive every vendor decision.
The deployable reference architecture is built open-source first, keeping the software foundation inspectable, portable, and operable outside a proprietary ZTA stack.
Approved models run in the institution’s controlled environment and can be re-benchmarked, replaced, or re-routed as the model landscape changes.
The architecture is designed to fit existing data, identity, security, workflow, and infrastructure choices rather than forcing a rip-and-replace decision.
Begin with Investment Research
The first production workflow for institutional investors, showing approved evidence, governed model analysis, Human Judgment, decision records, and measurable outcomes in one controlled Investment Research process.
Prepare committee materials using the same evidence, review, and decision-record discipline.
Apply the sovereign foundation to a governed diligence workflow with controlled evidence and review.
Track watchlists, emerging signals, and policy-driven escalations over time.
