No subscription · No paywall · We don't sell user data to pharma
Free cross-vendor NGS unification for pathologists.
Upload an NGS report from any lab. The parser handles Foundation, Tempus, Caris, Guardant, ARUP, LabCorp, and Natera by name, with a generic parser for everyone else and OCR fallback for scanned PDFs. Get a unified, citation-backed interpretation back. No paywall on usage. No subscription. We do not sell user data to pharma.
The free beta is live as a closed beta. Request access with your institutional email; we review clinical-use requests within five business days, and you can start uploading once you're approved.
Unified summary
Synthetic case- VariantEGFR p.Leu858Arg
- TierAMP/ASCO/CAP Tier I-A · FDA-approved
- TherapyOsimertinib (FLAURA)
- ContraindicationPD-1/PD-L1 monotherapy in EGFR+ NSCLC
- TrialNCT04988295 · MARIPOSA-2
Every claim links back to the source report, the FDA label, or ClinicalTrials.gov on the source's own site.
The problem
Five vendors. Five layouts. Same genomics.
On a typical week, a working pathologist or oncology fellow opens NGS reports from Foundation Medicine, Tempus, Caris, Guardant, and Natera. Each clinical report runs 10 to 30 pages, with technical appendices that can run longer. Each follows the lab's own layout. Each labels biomarkers a little differently. Each puts the actionable variants somewhere different on the page.
The clinical content overlaps significantly across vendors. The presentation does not. EGFR L858R is EGFR L858R wherever it shows up. The variant nomenclature, the evidence levels, the drug-gene match logic do not change between Foundation and Tempus. What changes is everything around the variant: the page layout, the order of sections, the way “Tier I” is shown versus “Level 1,” the color coding of contraindications, the location of the trial-match block.
So clinicians spend mental cycles on translation. Reading three vendor reports for a complex case takes longer than it should because the brain has to context-switch between layouts before reasoning about the actual clinical content. The high-cognitive-load part of this work is not the genomics. It is the formatting.
Engine 4 fixes the formatting. Upload the PDF, get a unified summary in the same structure regardless of which lab it came from. Every claim is cited back to the source report, the underlying knowledge base, or the FDA label. No fabricated citations. No invented drug names. No “this might be relevant” hand-waving.
How it works
Upload the PDF. Get a unified summary back.
Upload returns right away. The report is parsed in the background while the page polls for the result, so you can move on and come back to a finished, AMP/ASCO/CAP-tiered interpretation instead of watching a spinner. The pipeline behind it is the same one that powers the UNMIRI NGS Interpretation API, but the output here is rendered for a clinician's eyes rather than for downstream software ingestion.
Variants
Normalized to HGVS notation with transcript references, regardless of how the source vendor formatted them. EGFR p.Leu858Arg is shown the same way whether the report came from Foundation, Tempus, or Caris.
Biomarkers
TMB, MSI, HRD, and PD-L1 displayed in a consistent format with the lab-reported value, the test method (IHC clone for PD-L1, panel size for TMB), and the clinical interpretation.
Therapy options
Drug-gene matches with AMP/ASCO/CAP 2017 evidence tiers and FDA approval status. Same Tier I-A designation for the same drug-variant pair regardless of which lab originally reported it.
Contraindications
Drug-class flags from openFDA labels. The PD-1/PD-L1 inhibitor flag for EGFR-mutant NSCLC fires the same way whether the variant call came from a tissue or liquid biopsy report.
Trial matches
ClinicalTrials.gov matches with NCT IDs, eligibility text, and recruiting status, all linkable back to the source registry.
Lab-specific addenda
If a vendor included an interpretive note unique to their workflow, the summary preserves it as a quoted passage with the lab's name attached, rather than absorbing it into the unified output.
The tool is honest about what it finds. If a variant has no graph match, the summary says so rather than guessing. If an LLM ever contributes to extraction, the affected fields are flagged, and no citation is ever invented. It also recognizes report types that carry no DNA variants by design. Upload a gene-expression prognostic assay (Oncotype DX, Decipher, Prosigna, Afirma), a ctDNA or MRD monitoring test like Signatera, or an NIPT or carrier screen, and the tool names the assay and surfaces its headline result, such as a genomic- classifier risk category or a ctDNA-detected status, as a clearly labeled biomarker. It never presents a risk score as a variant interpretation. When a score is printed only inside a gauge and never in the report text, the tool still tells you which assay it is instead of guessing a number. A negative panel comes back as a clean negative result, not an error.
Vendors parsed by name today: Foundation Medicine, Tempus, Caris, Guardant, ARUP, LabCorp, and Natera. A generic parser handles any other lab's format, and image-based or scanned PDFs fall back to AWS Textract OCR. Named-vendor coverage is expanded as user demand justifies the parser work.
Why it's free
The freeness is structural, not a growth tactic.
The engine itself (parsers, knowledge graph, deterministic rendering) is shared across all four UNMIRI products. The marginal cost of running it for a pathologist is small. The strategic value of having practicing clinicians use it is large: each parsing edge case clinicians flag improves the engine that the paid API products run on.
Monetization at scale will come from pharma educational sponsorship, beginning once the user base passes a credible threshold. The internal target is 1,500 or more active monthly users, with a late 2027 timeline. Sponsorship is on educational content adjacent to the tool (landscape reviews, KOL panels, congress recaps), clearly disclosed and labeled. The summary output itself is never paid placement and never promotes a specific drug.
UNMIRI does not sell user data to pharma. We do not aggregate de-identified usage and resell it. The user base is the asset; selling access to that asset directly would burn the trust that makes the asset valuable in the first place.
What's the catch
There isn't one.
The tool is free for individual clinical use, the citations are real, the data is not sold, and the freemium-to-paywall trap that has soured pathologists on consumer-grade clinical tools does not apply here.
Two honest caveats. First, we ask for a work email at signup, institutional rather than personal, so we can verify clinical use and so the field force at our pharma sponsors does not end up surveilling the user base via signup data. Second, the tool is not a CAP/CLIA-validated diagnostic. Output is a decision-support aid for clinicians who already have the underlying validated lab reports in hand. Don't paste a summary into a sign-out without the source report's lab director cosignature in the loop.
Compliance posture for the underlying platform (subprocessor list, BAA status, region pinning) is on the security overview. Engine 4 is scoped to non-PHI clinical use by default; institutional deployments that want PHI scope follow the API-product path on Engines 1 and 2.
Roadmap
Three stages.
- 01
Free closed beta · live now
Access is open by request. Pathologists, oncology fellows, and genetic counselors can request access with an institutional email; we review clinical-use requests within five business days, then you can start uploading NGS PDFs. The parser handles Foundation, Tempus, Caris, Guardant, ARUP, LabCorp, and Natera by name, with a generic parser for any other lab and AWS Textract OCR fallback for scanned PDFs. Early users get a direct line to the engineering team for parser feedback and edge-case escalation while the stack is hardened against real-world report variability.
- 02
Coverage expansion · ongoing
Named-vendor parsers are added as user demand justifies the work, and biomarker and trial-matching depth grows alongside the Engine 1 and Engine 2 pipelines the tool draws on. The generic parser keeps any lab usable in the meantime.
- 03
General availability · Q4 2026 to Q1 2027
The product graduates from beta with documented uptime, latency, and accuracy targets, and stays free for individual clinical use indefinitely. Sponsorship monetization begins late 2027, once the user base passes the 1,500+ active monthly user threshold that makes educational sponsorship economics work for pharma counterparties.
The roadmap is honest about its dependencies. The biggest variable is parser coverage stability across vendor format changes. If a major vendor reformats during beta, that pushes the roadmap; we will not ship a tool that works inconsistently across last month's and this month's report layouts. If the parser stack stays stable, the timeline holds.
Who's building this
Two founders, software-first, with clinical advisors being recruited.
Engine 4 is part of UNMIRI, a precision oncology infrastructure company built by a two-person founding team: Nida Uddin (Founder and CEO, leading operations, finance, and commercial strategy) and Umair Khan (Founder and CTO, software architect with 14+ years of engineering experience). UNMIRI was started after Umair's caregiving experience exposed how fragmented the precision oncology software ecosystem is for clinicians who interact with it daily. The full founder story is at /about.
Clinical accuracy is verified through ongoing recruitment of board-certified pathologist advisors. Public introductions are added once each engagement is formalized and the advisor approves being named. The closed-beta cohort overlaps with this recruitment process, so early beta users have a direct line to both the engineering team and the clinical advisors as they come on.
Sample output
Best way to know if this is useful is to see it.
We've built a fully rendered sample for a synthetic NSCLC case with EGFR L858R, TP53 R175H, and PD-L1 <1%, including the contraindication flag for checkpoint-inhibitor monotherapy in EGFR-mutant disease and a pre-matched MARIPOSA-2 trial entry for the resistance pathway. Same output structure the beta tool produces.
View the sample reportMore to inspect
- /samples/compare: Foundation vs Tempus vs Caris
- EGFR L858R walkthrough: full case study
- EGFR cheat sheet: 2 minutes at the microscope
- VarSome comparison: different verb, different fit
Beta signup
Start the free beta.
One short form with an institutional email so we can verify clinical use. We approve requests within five business days, then you can upload right away.
One short form. Free for individual clinical use. Approved by review within five business days.