Explainable clinical-trial matching

Find clinical trials matched to your health profile

MatchVox analyzes your clinical information against trial requirements and produces a ranked shortlist with understandable reasons for every potential match.

Recruiting for a study? Explore MatchVox for institutions.

87.8M
Patient–trial pairs analyzed
520K+
Trials indexed
3s
Per patient
AUC 0.703
External gold-label evaluation

Two paths, one platform

Whether you're exploring treatment options or filling enrollment, start here.

Patients & caregivers

Discover potentially relevant trials and understand why each trial may or may not fit your health profile.

Share as much or as little of your health information as you like. Every ranked result comes with plain-language reasons, so you and your clinician can review the shortlist together.

Find matching trialsFree · Processed in your browser

Sponsors, CROs & research sites

Identify potential participants, reduce manual prescreening and give research teams reviewable eligibility evidence.

Run a focused recruitment pilot for one actively enrolling study — retrospective or de-identified data, criterion-level evidence for coordinator review, and measurable screening-time comparison. No production integration required for a first evaluation.

Request a pilot2–4 week evaluation · One protocol
HIPAA-ready
Externally evaluated (TREC 2021)
Published methodology
Evidence shown for every result

HOW IT WORKS

Three stages. Three seconds.

A two-stage retrieval-and-reranking architecture screens all 520,484 registered trials in about 3s per patient.

PATIENTDOMAIN-DRIVEN DEEP LEARNINGTREATMENT & CARE TEAM
01

Input

Your health profile is parsed and understood in full clinical context — diagnoses, labs, medications, history.

02

Intelligence

Semantic retrieval + precision reranking scores every trial in the database against your specific profile.

03

Output

Ranked results with plain-language reasons for each — including what supports a match, what conflicts, and what is unknown.

520K+
Trials indexed
3 sec
Per patient
5,418×
Faster than exhaustive pairwise scoring

The 5,418× figure compares our two-stage pipeline against exhaustively scoring every patient–trial pair with the cross-encoder — not against manual review or other products.

EVALUATION

The numbers, with their provenance

Every metric below names its dataset and label type. Internal metrics use silver labels; the external evaluation uses independent gold labels the model never saw. See our methodology and limitations.

AUC 0.703
External evaluation — independent gold labels
TREC 2021 Clinical Trials Track gold relevance labels; BM25 keyword baseline scores 0.472 on the same set (a 49% improvement).
0.900
NDCG@10 — internal evaluation, silver labels
Internal test split scored against silver labels (Jaccard similarity, τ = 0.020). Silver-label metrics share their labeling process with training and should be read alongside the external evaluation.
98%
Recall preserved by approximate retrieval
of relevant trials retained by the approximate first-stage retrieval, measured on internal evaluation sets.

Open Evaluation

Benchmark us yourself

Same patients. Same trials. Same criteria. We welcome head-to-head evaluations on shared datasets and will provide our evaluation harness so you can verify every number we publish.

Request a benchmark
5

Research papers behind the methodology.

One published paper and four papers currently in submission. Every claim on this site traces to a documented evaluation.

View our research
86%

of clinical trials fail to recruit on time

Every day without the right match is a day of treatment delayed.