Selected engagements.

A short, representative selection of the work we do — each showing how an engagement runs from problem to evidence.

Representative engagements. Client identities and specific figures are withheld; each illustrates the shape of an engagement and what confidence looks like. We do not publish results we cannot substantiate.

Finance & audit

Intercompany reconciliation for a multi-entity group

The situation

A group closing several entities every month reconciled intercompany balances by hand. Volume and entity count had outgrown the spreadsheet, and differences surfaced late in the close.

What we did

We trained a model on the group's chart of accounts and prior reconciliations, integrated it with the ERP, and routed unmatched items to owned exception queues with source-document links.

What confidence looked like

Each match traced to its documents and model decision; exceptions assigned; close evidence assembled as the work happened.

Pharmaceuticals

Safety signal review at scale

The situation

A pharmacovigilance team reviewed large volumes of safety reports with limited capacity, while regulatory scrutiny demanded a clear, defensible audit trail.

What we did

We built a traceable review workflow — model-assisted triage with record-level lineage and human sign-off on material signals — integrated into the existing safety systems.

What confidence looked like

Reproducible review, human oversight on material calls, and exportable evidence for inspection.

Genomics

Reproducible variant interpretation

The situation

A genomics team needed variant interpretation to be reproducible across runs and defensible in review — not dependent on an individual analyst's settings.

What we did

We made the pipeline versioned and deterministic, recorded reference and model versions, and tied each call to the reads behind it.

What confidence looked like

Same input, same output; every call traceable to its evidence and versions; records ready for peer or clinical review.

Mining & exploration

Assay QA/QC traceability for a drill programme

The situation

An exploration programme relied on assay data with QA/QC checks spread across files, making it hard to demonstrate confidence in the data feeding a resource estimate.

What we did

We integrated collar, survey and assay data into one traceable record, surfaced QA/QC issues early, and linked every result to its source sample.

What confidence looked like

Assays traceable through QA/QC; data-quality issues surfaced before modelling; records defensible to investors and regulators.

Your engagement could be next.

Bring us a real problem and a sample of your data, and we will show you how we would approach it.

[email protected]