Product & services

A simulation, and the evidence that it works.

AlphaIQ is not a dashboard you log into and trust. It is a modelling capability that produces four artifacts — specification, code, calibration, validation — each of which you own, read and can hand to a reviewer.

How we work

Three ways to start.

Model build

A single question, modelled end to end.

You bring a decision that a simulation could inform. We produce the specification, the calibrated model and the validation record, and we run the scenarios with your team rather than for them.

Best when there is a specific question and a date it has to be answered by.

Structured model specification, reviewed with your team
Executable model code, delivered to you
Calibration against your data or public data
Validation record: sensitivity, residuals, plausibility
Scenario runs and a working session on the results

Capability build

The pipeline, deployed in your environment.

For teams that will model repeatedly, we deploy the AlphaIQ pipeline into your cloud account or on-premises and work alongside your modellers until they are running it without us.

Best when simulation is becoming a recurring part of how a team works.

Pipeline deployed in your environment
Integration with your data sources
Model library seeded with your existing models
Training for your modelling team
Handover documentation and a support period

Model review

An independent read of a model you already have.

We take an existing model — yours or a vendor's — reconstruct its specification, re-run its calibration and produce the validation evidence that may not currently exist for it.

Best ahead of a review, a committee, or a decision to rely on someone else's model.

Reconstructed specification of the model as built
Independent calibration check
Sensitivity and failure-boundary analysis
A written assessment of where the model holds and where it stops

What you keep

Four artifacts, not a subscription.

The intent is that an AlphaIQ engagement leaves you less dependent, not more. Everything below is yours, in an open format, at the end of the work.

The specification

Agent types, state, behavioural rules and parameters in structured form — the object the code is generated from, and the thing a reviewer should argue with.

The model code

Executable simulation code in Python, readable and modifiable by your own team, with the test harness used to check agent behaviour in isolation.

The calibration record

Target data, distance measure, accepted parameter region, and an explicit list of what was fixed by assumption rather than fitted.

The validation record

Residual analysis, distribution comparison, sensitivity sweeps and plausibility checks — including the conditions under which the conclusion reverses.

Calibration record: posterior parameter regions and residual diagnostics — The calibration and validation record is part of the deliverable

Under the hood

A team of specialised agents, supervised.

The pipeline is a multi-agent workflow rather than a single prompt. Each agent has a narrow job and writes its output into a shared project record, so the progression from source document to validated result is traceable end to end.

Supervisor: Manages the workflow: interprets project state, sequences the work, routes tasks to the right specialist and handles failure.
Analyst: Reads source material — papers, specifications, documentation — and extracts model logic into a structured specification.
Developer: Translates the specification into executable model code, with automated review and iterative correction.
Data scientist: Prepares data, runs calibration and optimises parameters against observed behaviour.
Operator: Executes simulations at scale and abstracts away the compute infrastructure underneath.
Validator: Compares simulation output against empirical data, runs sensitivity analysis and interprets residuals.

Deployment is designed to run inside your own cloud account or on-premises. Describe your specific security and compliance requirements to us and we will tell you plainly whether we currently meet them.

AutoSim, our product Book a demo How it works