~20 min → <4 s
Per-PDF review time
Illustrative POC/MVP reference: manual review versus local-GPU assisted processing.
Case study
Forensic drug analysis — lab report review
An illustrative design-partner narrative for the Ivertiq Full-Stack AI Harness: LC–MS / LC–MS/MS instrument PDFs, sovereign on-prem execution, and human experts on every ambiguous or conflicting case. Not a guarantee of results in every lab.
2,000+
PDFs in a batch window
Reference scale on workstation-class dual-GPU hardware for monthly-style processing loads.
On-prem
Sovereign by design
Forensic evidence, Lab IDs, and chemical results stay inside the customer environment.
Metrics are illustrative POC/MVP selling aids pending validation under each customer’s SOP, QA protocol, input formats, and hardware.
The problem
In forensic drug analysis, LC–MS and LC–MS/MS (and in some workflows GC–MS) are the analytical gold standard: liquid or gas chromatography separates a complex sample into ingredients; mass spectrometry characterizes each ingredient and compares its spectrum against reference databases.
Even so, the instrument does not issue a finished judgment. For each ingredient it typically emits a measured spectrum plus database candidates — often HIT 1 and HIT 2 from two libraries — with similarity / confidence scores. Those scores are useful starting points, not sign-off. High-confidence candidates can still be wrong or ambiguous when peak patterns, impurities, noise, or retention-time context disagree. Labs therefore require expert visual inspection of the equipment-generated PDFs before findings are recorded and reported.
As volume rises (thousands of PDFs per month in some settings; tens of thousands of samples per year), adding headcount alone does not scale. Fatigue increases QC risk. Sensitive evidence, Lab IDs, and chemical results cannot be casually uploaded to public cloud AI.
Instrument confidence is not a final answer. The goal is not to replace forensic experts — it is to accelerate disciplined review, surface conflicts early, and keep every accountable decision human-gated.
Illustrative walkthrough
An illustrative forensic LC-MS proof walkthrough from one customer’s POC/MVP. Not a general production guarantee. Timing and hardware figures remain POC/MVP references pending each lab’s SOP, QA protocol, and hardware.
What makes the PDF hard
- Crowded evidence — each ingredient may present multiple spectrometric diagrams that must be read together with labels, retention time, and text HIT results.
- Close or conflicting HITs — two strong database matches can both be wrong, or only one may be correct once impurity peaks are excluded.
- Mandatory expert gate — similarity scores alone are not sufficient for sign-out; humans still certify the read under lab SOP / QA.
Phase 1 scope (illustrative)
01
Ingest
Instrument PDFs / exports and controlled mapping references.
02
Extract & map
Lab ID, HIT 1 / HIT 2 candidates, similarity cues, and spectrum-linked evidence → controlled drug mapping rules.
03
Gate
Consistent mapping → structured output; conflict, close HITs, or low-quality spectra → expert review.
04
Collate
Optional PPT assembly from LIMS export + selected sample photos.
Why it proves the Harness thesis
The same stack pattern that matters here — local execution, structured evidence, deterministic conflict gates, and HITL — is why generic public-cloud copilots are the wrong default for forensic PDFs.
- Sovereignty — forensic data stays local; no public LLM API path for evidence.
- Deterministic gates — automation stops when Hit1/Hit2 disagree; experts own exceptions.
- Auditability — structured outputs and review paths support defensibility.
- Expansion path — same stack pattern can transfer (discovery / method transfer) to clinical/health-screening LC-MS review and other lab document workflows. That is not a live module.
Forensic case FAQs
Questions specific to this proof case. Company, partner, and investor FAQs stay on the FAQ page.
What problem does this address?
It assists review of large volumes of LC–MS / LC–MS/MS instrument reports in forensic drug-analysis laboratories. It does not replace the laboratory’s SOP or the expert who signs out the case.
What types of laboratory reports can it process?
Instrument-output PDFs containing chromatograms, mass spectra, retention times, library hits, and related analytical evidence.
Why is manual LC–MS/MS review so demanding?
Analysts must interpret several evidence types together — peak patterns, retention time, signal quality, and database matches — before a finding is recorded.
Does the system replace forensic analysts?
No. Assist, not autopilot. The Harness helps read the PDF and surface conflicts. Uncertain or ambiguous findings go to a qualified reviewer. Accountable sign-out stays with a designated person.
How does the assist form a likely substance?
It evaluates database candidates together with peak coverage, retention-time consistency, noise or interference, and agreement between chart evidence and reported hits. A similarity score alone is not sign-off.
Why are two library matches sometimes different?
Different libraries, reference spectra, matching algorithms, instrument settings, and sample interference can produce different rankings for the same sample.
What happens when the evidence is unclear?
The system flags cases with close candidates, missing peaks, high noise, weak hit agreement, or unusual retention times for expert review. It does not guess-merge or pick a “closest” substance.
Can it handle impure or complex samples?
Yes. The workflow is designed to assess interference, background noise, and overlapping peaks rather than relying on a single similarity score.
Is the reasoning reviewable?
Yes. Each recommendation includes structured evidence and metrics that link back to the relevant source-report pages, diagrams, and extracted data.
Does sensitive case data leave the laboratory?
Core evidence stays in a customer-approved environment — on-prem Node/Station or siloed private cloud. Sensitive reports and identifiers are not sent to public LLM APIs. Air-gap is a real path where the site requires it.
How does this support auditability?
It keeps a reconstructable trail: evidence, scores, actions, and review outcomes for each result — attributable to a model, recipe, skill, or named person.
What is human-in-the-loop (HITL) review?
A designated person remains accountable. They can verify, correct, or approve before a result is finalized. Nothing of legal or scientific effect writes until that explicit yes.
How quickly can reports be processed?
In the illustrated proof of concept, average assist time fell from roughly 20 minutes per PDF to about four seconds per PDF. Actual performance depends on workflow, hardware, and validation. It is not a guarantee for every lab.
Can the platform scale as laboratory volume grows?
Capacity scales with local compute. Larger batches can run on customer-approved hardware. Staffing and sign-out remain the laboratory’s SOP decision — we do not claim a headcount replacement figure.
Is this only for forensic drug analysis?
Forensic LC–MS / LC–MS/MS PDF review is the current proof. The same method may transfer to other regulated lab document workflows when the SOP, HITL gate, and system of record are named. Clinical chemistry, environmental testing, and food safety are discoveries — not live modules.
How is this different from a generic chatbot or public-cloud assistant?
Local or private-cloud deployment, deterministic workflow steps, domain-specific validation, human-review gates, and audit logging. It is not a chat overlay on a public LLM API.
What we do not claim
- Replacement of forensic expert judgment
- Guaranteed accuracy or elimination of all QC risk
- Coverage of every forensic subsidiary workflow on day one