Perceive
Reads any lab report — PDF, photo, or text — and turns messy clinical documents into structured, normalized biomarker data. Units reconciled, values mapped, nothing dropped.
OCR + Claude Haiku 4.5 · $0.001 / patient · ZDR + BAA
Four audiences
AI for health, deployed for doctors, patients, employers and researchers.
Not a chatbot. A clinical reasoning pipeline that perceives, knows, and reasons — the same three layers a physician uses, running on validated medical knowledge.
A doctor has minutes per patient and can't hold every study and every cross-marker interaction in their head. The core can. It reads the full profile every time — then the same mind is delivered four different ways.
The core
Reads any lab report — PDF, photo, or text — and turns messy clinical documents into structured, normalized biomarker data. Units reconciled, values mapped, nothing dropped.
OCR + Claude Haiku 4.5 · $0.001 / patient · ZDR + BAA
102 biomarkers with reference ranges specific to age and sex, plus a dependency map of how markers relate. Built from peer-reviewed literature, checked against population data.
102 biomarkers · age/sex ranges · HOMA-IR + BioAge
15 clinical patterns and 8 types of interaction between them. The engine sees the whole picture — how inflammation, insulin resistance and lipids connect — not isolated out-of-range numbers.
15 patterns · 8 interaction types · F1 0.963 best-in-benchmark
Engine map
Every product is a different assembly of the same building blocks. Select any node to see what it does and where it's reused — that reuse is the whole point.
Reads PDFs, photos or text and turns messy lab reports into structured, unit-reconciled biomarker data.
Engine map
Every product is a different assembly of the same building blocks. Select any node to see what it does and where it's reused — that reuse is the whole point.
Age- and sex-specific reference ranges plus a dependency map of how markers relate.
Engine map
Every product is a different assembly of the same building blocks. Select any node to see what it does and where it's reused — that reuse is the whole point.
Detects 15 clinical patterns and 8 interaction types across the full profile. F1 0.963 best-in-benchmark.
Engine map
Every product is a different assembly of the same building blocks. Select any node to see what it does and where it's reused — that reuse is the whole point.
Derived indices computed from the normalized profile.
Engine map
Every product is a different assembly of the same building blocks. Select any node to see what it does and where it's reused — that reuse is the whole point.
Peer-reviewed literature retrieval that grounds every synthesis.
Engine map
Every product is a different assembly of the same building blocks. Select any node to see what it does and where it's reused — that reuse is the whole point.
Zero-data-retention inference with signed BAAs; HIPAA-aware by default.
Engine map
Every product is a different assembly of the same building blocks. Select any node to see what it does and where it's reused — that reuse is the whole point.
Server-side enforcement so employer dashboards never expose an individual.
Engine map
Every product is a different assembly of the same building blocks. Select any node to see what it does and where it's reused — that reuse is the whole point.
Longitudinal symptom capture linked to labs and medications.
Engine map
Every product is a different assembly of the same building blocks. Select any node to see what it does and where it's reused — that reuse is the whole point.
Treatment schedule the AI reads as part of the anamnesis.
Engine map
Every product is a different assembly of the same building blocks. Select any node to see what it does and where it's reused — that reuse is the whole point.
Shared, permissioned view for family and caregivers.
Engine map
Every product is a different assembly of the same building blocks. Select any node to see what it does and where it's reused — that reuse is the whole point.
Population datasets the pipeline runs on and the engine is validated against.
Engine map
Every product is a different assembly of the same building blocks. Select any node to see what it does and where it's reused — that reuse is the whole point.
Automated analysis with adversarial hypothesis generation.
Engine map
Every product is a different assembly of the same building blocks. Select any node to see what it does and where it's reused — that reuse is the whole point.
Autonomous scan of current literature to seed hypotheses.
Engine map
Every product is a different assembly of the same building blocks. Select any node to see what it does and where it's reused — that reuse is the whole point.
Five independent models review every draft before publication.
Engine map
Every product is a different assembly of the same building blocks. Select any node to see what it does and where it's reused — that reuse is the whole point.
Produces a preprint-ready manuscript from the analysis.
Zenlo
“Like a doctor who has read everything and forgets nothing, without the time limit.”
One clinical AI mind · built by an engineer · validated in the open
Products
Status badges are honest — they reflect where each product is today, not the maturity of the engine underneath.
Upload a lab PDF, get a physician-ready synthesis — patterns, HOMA-IR, biological age, an action plan you can sign and lock. A private clinical workspace; the doctor decides what to share.
Preview Labs →Early detection of chronic-disease risk across a workforce — without ever exposing individual health data. Aggregate-only dashboards with k-anonymity enforced server-side. $2 PMPM pilot.
Request a pilot →Full-profile recovery support — labs, medications, symptom journal, therapy calendar, and an AI that knows the whole anamnesis. Caregiver access built in. Closed pilot. Not diagnostic.
Join the early list →An autonomous pipeline: literature scan → adversarial hypotheses → statistical analysis → 5-model clinical review → draft paper. Runs on NHANES + BRFSS. Preprints published in the open.
Read the papers →Latest from Zenlo
Every claim on this page traces to a paper or a public benchmark — no marketing numbers.