built by an engineer · validated in the open

One mind for the whole patient.

A clinician gets fifteen minutes, and the evidence that should inform the visit is scattered across more research than anyone can read. Zenlo reads the whole profile against the whole literature — and hands the synthesis back to the doctor.

Medicine doesn't have a knowledge problem. It has a time problem.

The study that would change a decision usually already exists — it's just buried in a literature no clinician can keep up with, waiting to be applied in a visit too short to apply it. Meanwhile a patient's own data — labs, history, medications, trends across years — rarely gets read as one connected picture.

Chronic disease is now 86% of U.S. healthcare spending, and the gap between catching something early and catching it late can run 26× in cost. A large part of that gap is simply reading that no one has the hours for. Zenlo is the reading layer — built to do, every time, what a doctor would do if they had read everything and forgotten nothing.

Zenlo proposes. The clinician decides. Every synthesis is a draft until a doctor reviews it, edits it, and signs it. We surface the patterns and the things worth a second look — then we stop, exactly where clinical judgment begins. We never deliver a conclusion around the physician. That isn't caution bolted on afterward; it's the shape of the product.

A lab report goes in. A physician-ready synthesis comes back.

One shared clinical core does the work — the same engine behind every Zenlo product. Here is what happens in between.

01 / Extraction

Document to data

A PDF, a photo, or raw text becomes structured biomarker data — normalized across units and reference ranges by age and sex.

Claude Haiku 4.5 · ~$0.001 / patient
02 / Pattern recognition

Patterns, not single markers

15 validated clinical patterns, plus interaction analysis across eight types of relationship between them — because risk rarely lives in one number.

F1 0.963 best-in-benchmark
03 / Quantified signals

Indices that mean something

Insulin resistance (HOMA-IR), biological age, and longitudinal trends — computed and explained, not left as raw figures.

HOMA-IR · BioAge · Trend Engine
04 / Evidence grounding

Checked against the literature

Every read is grounded in a knowledge store of 2,534 documents spanning population data and published research, 2000–2025.

pgvector · NHANES + PubMed
05 / Memory

The full picture, carried forward

Zenlo holds the patient's standing history, evolving results, the relevant literature, and the tools that connect them — as one context.

static + dynamic + retrieval + tools
→ Go deeper

See the engine, block by block

The full architecture — every block and how the products reuse them — lives on the engine map.

Three principles, not slogans.

Principle 01

Doctor-in-the-loop

The synthesis is a draft until a clinician signs it. Zenlo informs the decision; it never replaces the person making it. This is also why we stay on the right side of the line that separates clinical software from a medical device.

Principle 02

Honest by default

We label what's live and what we're still building, on every product. No demo-ware dressed up as shipped product. We would rather show you a smaller true thing than a bigger false one.

Principle 03

Privacy is the architecture

Zero-data-retention AI under signed business-associate agreements, HIPAA-aligned design, row-level isolation per user, and k-anonymity (k ≥ 10) on anything aggregated. A file is the patient's; the doctor decides what is ever shared.

We don't ask you to take accuracy on faith.

Zenlo's pattern engine was benchmarked across five frontier models and tested against 4,018 NHANES patients. The methods — and our audits — are published.

Preprint Feb 2026

A five-model benchmark for AI clinical biomarker interpretation

medRxiv · MEDRXIV/2026/346284 · CC-BY 4.0
In peer review 2026

Multi-model validation of automated biomarker pattern detection

JAMIA Open · manuscript under review
Self-published 2026

Security & code audit — Zenlo Labs

labs.zenlo.app/audits · methodology open

Deliberately small. Shipping in sequence.

Founder

Dmitrii Shibakov

Engineer and founder, building Zenlo end to end from Los Angeles — from the clinical engine to the product you use.

Medicine

Clinical advisory board

Practising clinicians guide the medicine, the patterns, and the boundaries — so the engineering stays accountable to clinical reality.

How we operate

Bootstrapped

No outside capital, no pressure to overstate. One solid product at a time, validated before it goes out.

The same mind, ready when you are.