Release candidate
Claim Audit Lab
A claim-level assessment environment for applying consistent support verdicts, preserving reviewer notes, and calibrating automated evaluations against human judgment.
About
Record-chain discipline for any workflow that has to stand up to review.
My training ground is regulated pharmaceutical quality. The useful part is portable: when an output can be questioned by a buyer, auditor, partner, or internal reviewer, someone still has to show what entered, what changed, who decided, and what was retained. That holds in life sciences, software, and service operations.
Background
Built from the record chain outward. Not from a single industry slogan.
I hold an Honours B.Sc. in Biological Science and have professional experience spanning sterile allergy-immunotherapy compounding, pharmaceutical and cosmetic laboratory testing, controlled documentation, quality investigations, data integrity, SOP development, method implementation, and work under cGMP, GDP, GLP, USP, and Health Canada-facing quality environments.
That background is not a claim that every engagement is pharmaceutical. It is why I default to provenance, intended use, version, exceptions, and human review when AI or automation touches a consequential workflow. The same failure modes show up outside the cleanroom. A confident output with a weak trail is still a weak trail in batch documentation, AI-assisted review, software delivery, service operations, or any other path someone has to defend later.
I build tools and methods that keep evidence, inference, generated text, and human judgment distinct, then apply that discipline in client and collaborative work across regulated and non-regulated settings. Where data governance, log integrity, and auditability determine whether a system is trustworthy, the industry label matters less than the record chain.
I work independently from Canada, remotely with organizations and collaborators in other markets, and am bilingual in English and French.
Selected systems
A portfolio organized around evidence reliability
These projects are separate components with explicit boundaries. Together, they support research, delivery, and future product development across domains where claims and workflows must remain inspectable.
Operational component
Evidence Bundler
A provenance-preserving component for nominating candidate passages and sealing evidence bundles without confusing retrieval coverage with support determination.
Active research
Research Scaffold Harness
Experimental infrastructure for testing whether structured research controls reduce unsupported claims without winning by making answers empty or excessively cautious.
Why this combination matters
Domain experience supplies judgment. The product is the control of evidence.
Quality work teaches that records, procedures, controls, and intended use matter. Software work teaches that a repeatable method can become a tool, workflow, or product. Research work adds another discipline: conclusions must remain proportionate to the evidence and uncertainty. Together, that combination travels: pharma is a high bar, not a ceiling.
Current practice
The primary productized offer is the Workflow Evidence Hardening Design Sprint, entered through a free one-workflow diagnostic. Controlled automation is a separately scoped follow-on when build is justified. Claims audits and evidence research remain available as secondary work when the problem is not the primary workflow-evidence shape. Fit turns on whether a workflow’s outputs must be reconstructed or defended, not on industry label alone.
Working on a workflow that needs a stronger evidence chain?
Start with one workflow and the audience expected to rely on it. Life sciences is one setting among several.