ZTA Labs

Institution-controlled AI for asset managers and asset owners

Your data.
Your models.
Your intelligence.

ZTA Labs helps asset managers and asset owners build and govern institution-controlled AI for Investment Research and other investment workflows, while retaining control of their data, models, infrastructure, agents, and institutional intelligence.

Talk to ZTA Labs
DataModelsComputeAgentsIntelligence
Built for institutional investors
Asset Managers · Hedge Funds · Pension Funds · Insurance · Reinsurance

The model is replaceable. The institution’s intelligence is not.

Why the architecture must change

AI is moving from assistants to agents. The architecture has to change with it.

As models and agents become embedded in Investment Research and other institutional investment workflows, asset managers and asset owners need control over what intelligence can access, which models can act, how much autonomy they have, and when Human Judgment is required.

01

Economics

Agentic AI changes the cost curve. Agentic workflows multiply model calls, retrieval, evaluation, retries, and tool use. Consumption can scale faster than the number of users.

Control the capacity, not just the token price.

02

Intelligence

Generic models do not accumulate your institution’s intelligence. A model does not inherently preserve research history, methodology, evaluation standards, corrections, prior decisions, or Human Judgment.

Your knowledge should survive the model.

03

Trust & governance

No single model is consistently reliable across institutional work. Higher-consequence use requires independent challenge, evidence, controls, and accountable Human Judgment.

Govern the decision, not just the model.

Data sovereignty is foundational to all three. Data, models, compute, entitlements, and the audit trail all need to remain governable inside the institution.

What ZTA delivers

A Production-Ready Sovereign AI Foundation.

ZTA Labs designs, builds, and deploys an open, client-controlled reference architecture for asset managers and asset owners. Built on open-source software and open-weight models, it supports private inference, governed data access, model orchestration, policy enforcement, Model Council, auditability, and Human Judgment.

  • Open-source first.
  • Vendor-flexible by design.
  • Institution-owned by default.
Open architecture
Open-source software
Open-weight models
Institution-owned intelligence

Client environment

Applications · Identity · Approved Data

ZTA AI Control Plane

Data access

What intelligence may access.

Identity & authority

Who, or what, may act.

Policy & risk

How much autonomy is permitted.

Model routing

Which intelligence should act.

Human control

What becomes a decision.

Model Council when required.

Client-controlled AI infrastructure

Open-source software · Approved open-weight models · Inference · Knowledge · Logging

Deployment foundation

Institution-selected infrastructure posture

What the client owns

Architecture pattern, software configuration, approved model registry, policy-routing logic, capacity model, security controls, logging and trace framework, deployment documentation, runbooks, and the ability to replace models, software components, infrastructure, or operating partners over time.

Business impact

Convert growing AI consumption into controlled capacity while making multi-model governance and continuous institutional learning economically scalable.

Model Assurance

Make model selection evidence-led.

ZTA independently benchmarks open-weight models against the institution’s actual Investment Research questions, documents, data, and evaluation standards, then maintains the evidence and governance framework required to determine which models remain fit for which investment tasks as models and institutional requirements change.

Prove model performance before production. Keep proving it after deployment.

Baseline qualification

Establish the governed corpus, benchmark the candidate pool, and determine which models are admissible.

Monthly assurance

Track material model, benchmark, or operating developments relevant to the approved pool.

Quarterly scorecard

Reassess the pool and update recommendations based on the evidence.

Annual governance audit

Review whether the methodology, controls, independence, and audit trail remain fit for purpose.

Event-triggered revalidation

Re-run when a new release, new use case, or material model change warrants reassessment.

Sample Model Assurance Baseline

See the methodology in practice.

Seven open-weight models evaluated across a governed institutional research corpus, with model-level findings, failure analysis, source-level auditability, governance requirements, and deployment recommendations.

View the Sample Model Assurance Baseline
100Questions
7Models evaluated
700Scored responses
3Models admitted to Model Council
Model Assurance Baseline — candidate evaluation100 questions · 7 candidates · 700 scored responsesEvidence verified
CandidateOverallHard failsModel Council
gpt-oss 120B96.41Admitted
Qwen3 30B A3B89.73Admitted
Mistral Small 3.1 24B87.24Admitted
DeepSeek-R1-Distill-Qwen 32B84.16Conditional

Admission is a standard, not a ranking. Hard-fail rate at or below 5% admitted, at or below 8% conditional, above 8% excluded. Illustrative engagement.

Qwen3 30B A3BAnalysis tier · sample baselineHard fail
Question

What was gross margin on a GAAP basis in the quarter?

Model answered

“50.4 percent” with a citation that resolves correctly to the earnings release.

Correct answer

50.3 percent. 50.4 percent is the non-GAAP figure three lines below on the same page.

Finding

Correct source. Wrong accounting basis. The error survives casual review precisely because the citation is correct.

The benchmark stays with the institution.

Investment Research · Research Management

Governed AI, inside the investment workflow.

Research Management is ZTA Labs’ first production workflow for institutional investors, designed around how asset managers and asset owners conduct, challenge, document, and learn from Investment Research. Approved evidence enters a controlled decision environment, risk policy determines the level of oversight, approved models analyze independently, and Human Judgment determines the institutional outcome.

Research Management · Credit deterioration assessment
Risk Tier 3 · Elevated · Model Council required

Assessment

Does recent financial and operating evidence warrant changing the internal credit view from Stable to Deteriorating?

Evidence set: 14 approved sources.

Model Council

Response A · 4.10

Evidence supports a cautious view but stops short of a rating change.

Response B · 4.42

Greater weight to deteriorating liquidity and covenant headroom, with a clearer downside-first framing.

Response C · 4.55

Evaluator recommendation. Strong on evidence coverage, less aligned to the firm’s risk posture.

Response D · 3.91

Underweights the change in operating flexibility and overstates the stabilizing factors.

The evaluator recommends. The investment professional decides. Institutional knowledge compounds.

Explore Research Management

Learning that stays with the institution, not the LLM provider.

Turn governed work into institutional learning.

Every governed workflow can generate institutional signal: evidence, model responses, evaluation results, Human Judgment, corrections, rationale, and outcomes. ZTA preserves that signal in a governed learning record so validated patterns can improve the institution’s AI capability over time.

01 · Governed work

Evidence, models, Human Judgment, and outcomes are captured from real institutional workflows.

02 · Capture

Responses, scores, corrections, rationale, traces, and outcomes enter the governed learning record immediately.

03 · Validate

Repeated observations, outcome confirmation, temporal relevance, consistency, and risk thresholds determine what is durable.

04 · Promote

Only validated patterns influence approved institutional memory, retrieval, routing, rubrics, or workflow behavior.

05 · Improve

The next governed workflow benefits from better evidence use, better routing, and a learning record that remains under institutional control.

Capture immediately. Validate before promotion. Improve continuously. Human Judgment is captured immediately. It becomes institutional learning only after validation.

Open by design. Owned by the institution.

Your AI capability should survive every vendor decision.

ZTA is designed to minimize structural vendor lock-in. Open-source software, open-weight models, documented interfaces, and transferable operating artifacts allow the institution to change models, inference engines, infrastructure, software components, or operating partners without surrendering its governance, decision history, learning record, or institutional intelligence.

We build it with you. You own it. You choose who operates it.

Replaceable technology

  • Open-weight models
  • Inference engines
  • Retrieval and knowledge components
  • Data platforms and integrations
  • Infrastructure
  • Operating and implementation partners

Institution-owned capability

  • Data and entitlements
  • Policies and governance logic
  • Benchmark corpus and evaluation methodology
  • Decision records and Human Judgment
  • Institutional learning record
  • Deployment documentation and runbooks

ZTA can design, build, deploy, and support the environment. The institution can operate and extend it internally or with another qualified partner. The architecture is designed to remain portable as models, software, infrastructure, and providers change. What your institution has learned should not.

Next step

Build AI capability your institution can own.

Start with Model Assurance, a governed Research Management workflow, or a broader sovereign AI foundation.

Talk to ZTA Labs