Skip to content
Applied AI engineering and research
IQAI
Evidence · engineering · human judgment
MenuClose

Intelligence systems for consequential problems.

Applied AI engineering

IQAI builds software for complex investigations, operations and decisions where AI, evidence and human judgment need to work together.

Explore systems → How IQAI works
Between AI capability and accountable work

AI for analysis. Rules for control.

IQAI is being built as control infrastructure between AI capability and accountable work. Evidence, explicit checks, review and human authorization help determine what can be relied on, rather than simply what a model can produce.

Capability

Models, data & tools

Produce material for analysis.

IQAI control infrastructure

Make reliance inspectable.

Evidence · checks · review

Required gate

Human authorization

Keep judgment, approval and accountability explicit before a result moves into use.

Accountable work

Decisions, records & operations

Use within an explicitly reviewed scope.

An architectural principle, not an automatic guarantee of correctness. Explore Artificial Brain
/01
Ground truth where available.A second AI opinion is not verification. Compare with independently verified evidence where available.
/02
Explicit rules where applicable.Use calculations, evidence gates, thresholds and policy checks for machine-checkable questions.
/03
Human authority where consequential.Keep judgment, approval and accountability explicit before a result moves into use.

Retained assessment / public extract · 24 September 2026

Pump running + zero flow

Candidate condition, not an established finding.

0 / 5 sufficiency requirements established

NOT EVALUATED

View retained assessment

Current systems

/01

Ground Truth

Evaluate tested AI outputs against declared independent, deterministic references. Keep results bounded to the tested tasks.
/02

Diagnostics

Inspect repeated responses, reflection changes and available token signals. Stability does not establish correctness.
/03

TRACE

Reconstruct people, organizations, money and events, with evidence behind each relationship.
/04

Reliance — Document Assurance

Check document claims against supporting material and record what still needs review.
/05

Spatial

Connect place, infrastructure and observations. Keep scenarios separate from recorded conditions.
/06

Gap Analysis

Turn captured operational evidence into explicit unknowns, targeted evidence requests and reviewable decisions.
Engineering approach

How the parts work together.

Start with the problem. Combine AI, rules, software and specialist methods, then test what the result can support.

Inputs · context · composition IQAI / living system Outcomes · learning
Problem · evidence · expertise · constraints
AIRulesIQAISoftwareMethods
Tested capability · reusable learning
IQAI living system Real-world problems, evidence, expertise and constraints move into IQAI engineering, which composes AI, rules, software and methods into working systems and reusable capability while preserving evidence, provenance and learning. Problem Evidence Domain expertise Constraints AI Rules IQAI Software Methods Working system Reusable capability Evidence Provenance Learning REAL-WORLD INPUTS COMPOSE DELIVER IMPACT
01UnderstandReal problems and context.
02ComposeAI, rules, software and methods.
03DeliverWorking capability in context.
04LearnReuse what proves useful.
Start with the operating question, available evidence, people and constraints. A model is not a substitute for access to the problem.

Explore IQAI

Systems

Systems built around different real-world operating problems.

How we work

Start with the problem, compose the right capabilities, validate, deploy and learn.

Team & community

Engineering, assurance and specialist experience close to the problems.

Research

Reliability, evidence, provenance, architecture and institutional use.

Built for the next decade, not the next model.

IQAI is being built to keep important problems, expert knowledge and new technical capability connected as the frontier changes.

Contact IQAI