# Camry Labs — capability brief

> Camry Labs enriches healthcare data with clinical annotation, expert evaluation of model outputs and review of existing labels, for AI labs and research teams directly and for annotation and data companies within their deliveries. It is owned and led by clinicians. Healthcare annotation is its field; in-house radiology, with radiologists working together in person, is its core today.

Revised 25 October 2026. This is the plain-text version of https://camrylabs.ai/capabilities; the facts here are the same as on https://camrylabs.ai/. Contact: contact@camrylabs.ai.

## Facts you can rely on

- Owned and led by clinicians.
- Core radiologists work together in person, in one office.
- Every annotation is reviewed by our in-house radiologists before delivery.
- Every engagement carries a clinician’s sign-off.

## What Camry Labs provides

Three services, each named with its plain meaning. None is model training, clinical validation or regulatory approval.

### Train — annotated training data

You need labelled medical data. Clinicians annotate to your label definitions and examples. For imaging, our in-house radiologists do the work together in one room, and every radiology annotation is reviewed before delivery.

- You send: The data, your label definitions and examples of what good looks like.
- You receive: Labels in your format, with the review status of every item.

### Evaluate — expert review of model outputs

You need clinicians to judge what a model says. Clinicians assess each output against criteria agreed with you: what is supported, what is wrong, what is missing, and why. Difficult cases are discussed and the reasoning is written down.

- You send: The outputs, and the questions you need answered about them.
- You receive: A review of each response, with reasons for the assessment.

### Validate — review of existing labels

You have labels and need them checked. Radiologists review labels you already have against the instructions they were made under, correct what needs correcting, flag what is uncertain and leave the rest alone, with the reasons recorded.

- You send: The dataset and the instructions it was made under.
- You receive: Corrections, the reasons for them, and what was left unchanged.

### Imaging work

Medical imaging is where the lab is deepest today: Ultrasound, CT, MRI, X-ray. The anatomy, the task and the review criteria are agreed before work begins.

- **Segmentation and delineation.** Contours, masks and region labels, in 2D images and 3D volumes.
- **Detection and classification.** Finding locations and structured labels, at image, series or study level.
- **Landmarks and structured labels.** Anatomical points, coordinates and attributes, to an agreed schema.
- **Expert annotation review.** A radiologist’s assessment of annotations you already have.

### Pilots and research collaborations

Work that can start before an annotation brief exists, with the clinical question agreed first. Each is an offer to find out together, not a claim of past results.

- **New imaging, new anatomy.** Anatomical annotation and segmentation review for volumetric ultrasound, whole-body and body-composition imaging, and other new imaging technologies.
- **Synthetic images, read by radiologists.** Anatomical critique of synthetic medical images: what is plausible, what is inconsistent, and what needs further investigation.
- **Where models fail, and why.** Difficult-case evaluation and analysis of model failures, with each of a model’s statements linked to the image evidence for or against it.
- **Where expert readers disagree.** Studies of uncertainty and disagreement: independent initial readings, recorded before any discussion, and the team discussion recorded separately.
- **Ultrasound, frame by frame.** Frame selection, view annotation and acquisition-quality assessment, so a model learns from the frames a sonographer would keep.

### Research support

Research groups engage the lab for study-specific annotation, expert image review and assessment of AI outputs, to the study protocol.

## Who Camry Labs works with

- **Directly — healthcare ai teams and research groups.** You bring the model, the data or the study. We agree the task and the clinical scope with you, do the work, and deliver in your format with a clinician’s sign-off. There is no intermediary.
- **Inside your delivery — annotation and data companies.** You keep your platform, your project workflow and your customer. We join as the clinical part of your team, work to your project instructions and review criteria, and take responsibility for what we deliver.

Behind both:

- **The same room.** Our radiologists annotate and review together, in person, in our office.
- **The same review.** Every radiology annotation is reviewed by our in-house radiologists before delivery.
- **The same signature.** Every engagement carries a clinician’s sign-off.

Platforms and tools members of the team have worked on: iMerit, V7 Darwin, RedBrick AI, Frekil, 3D Slicer. These are individual experience, not partnerships. The lab can work in your platform, a partner environment or a workflow arranged for the project.

## What is distinctive

- **Shared understanding..** One room, one reading of the instructions. When a definition is unclear, it is settled once, out loud, and recorded for everyone.
- **Difficult cases, discussed..** Difficult cases are raised for discussion as the work progresses. Where a project needs independent readings, we record them before any discussion: sitting together does not make a reading independent, so we do not claim that it does.
- **Accountability you can name..** Radiologists take part from the first conversation to the sign-off. Qualified engineers and technicians look after the tools and the working environment; the clinical work stays with the clinicians.

Clinical direction and quality oversight: Dr Yash Jakhotia, Radiology, MBBS, MD, DNB.

## How an engagement starts

1. Write to contact@camrylabs.ai, or send a brief through the form at https://camrylabs.ai/#contact. Say:
   - The kind of data and, for imaging, the modality and anatomy.
   - The task: annotated data, expert review of model outputs, a check of labels you already have, or a pilot on something newer.
   - Your label definitions, examples or review criteria, if they exist yet.
   - Volume, timing, and the platform you work in.
   - Whether you are engaging us directly or bringing us into a delivery.
2. Your brief is saved for our team. We reply to the email address you give us. A clinician joins the first conversation about scope.
3. Scope is agreed in writing: modality, anatomy, intended use, label schema, examples, exclusions, volume, working platform, output requirements and commercial terms, with a named client contact and clinical lead.
4. Where useful, a pilot on a representative sample, including difficult cases, reviewed together before scaling. Sample size, price and schedule are set per project.

Before any project data is shared, three things are settled in writing:

- **Access and workspace.** Who needs access, where the work happens, and whether anything may be downloaded or copied locally.
- **Transfer, storage and deletion.** How data moves, where it is stored, how it is protected, how long it is kept and who deletes it.
- **Confidentiality and permitted use.** The agreements your project requires, who owns the data and the labels, and what each may be used for.

The website is for project briefs only. It has no patient-data upload, and patient images or identifiable health information must never be sent through it.

### What a delivery can include

- **The data, in your format.** Labels, masks, coordinates or judgements in the schema we agreed.
- **Review status.** Which items were reviewed, corrected, or returned flagged as uncertain.
- **Decision log.** The rulings on hard cases, and the instruction version each applies to.
- **Correction record.** What changed after review, and why.
- **Release note.** What was checked against the agreed acceptance criteria, and who approved it.

## Questions buyers ask

**Can we engage you directly, or do we need to come through an annotation company?**

Either. Healthcare AI teams and research groups engage us directly. Annotation and data companies bring us into their delivery as the clinical part of their team. In both cases our in-house radiologists do the clinical work and a clinician signs off what we deliver.

**How do you work with data and annotation companies?**

You coordinate the wider project and keep your platform and your customer relationship. Our in-house radiologists provide specialist annotation and review within the responsibilities we agree, following your project instructions, review criteria and delivery format.

**How does an engagement start?**

Write to contact@camrylabs.ai or send a brief through the form on the homepage: the kind of data, the task, your label definitions or review criteria, volume, platform and timing. Your brief is saved for our team, we reply by email, and a clinician joins the first conversation about scope. A pilot on a representative sample usually follows.

**Can we start with a small pilot?**

Yes. We can scope a representative sample to align on instructions, review expectations and delivery format. Sample size, timing and price depend on the task and are agreed before work begins.

**Who does the work, and who reviews it?**

Our in-house radiologists lead annotation and clinical review, working together in the same office. Every radiology annotation is reviewed by our in-house radiologists before delivery, and every engagement carries a clinician’s sign-off. Dr Yash Jakhotia provides clinical direction and quality oversight. Qualified engineers and technicians support the tools and the working environment.

**Do you train or validate models?**

No. We make and review the data. Train means annotated training data, Evaluate means expert review of model outputs, and Validate means review of labels you already have. None of these is regulatory approval or a statement that a model is fit for clinical use.

**What kinds of healthcare data do you take on?**

Medical imaging is where the lab is deepest today: ultrasound, CT, MRI and X-ray. We agree the anatomy, the annotation task and the review criteria with you before work begins. For other kinds of healthcare data, tell us what your project needs and we will say plainly whether we can do it well.

**Our imaging is new. Can you work on it before there is an annotation brief?**

Yes, as a scoped pilot or a research collaboration. We agree the clinical question first, work through a representative sample with you, and write down what we learn, including where our readers disagree. What we deliver is the record of that work, not a claim about your technology.

**Can you work in our existing platform?**

Yes. We can work in your platform, a partner environment or a workflow arranged for the project. Access, tooling and responsibilities are agreed during scoping.

**Which output formats can we request?**

Tell us the format and schema your workflow needs. We confirm compatibility with the chosen platform and agree the delivery contents before production. A sample export can be checked first.

**What determines price, timing and minimum scope?**

The kind of data, the complexity of the task, volume, review requirements and the working platform. A representative sample helps establish a practical plan and quote. Pilot size, minimum scope and milestones are agreed for each engagement.

**Do you hold security or quality certifications?**

We do not claim any certification on this website. Access, transfer, storage, deletion, confidentiality and permitted use are agreed in writing for each project before any data is shared, and we say plainly which controls we can and cannot provide.

**Who owns the data and the final output?**

Ownership, permitted use, delivery rights and any restrictions on reuse are defined in the project agreement before work begins.

## What this brief does not claim

- No accuracy, agreement or quality figures: none are published here.
- No turnaround or capacity commitments: timing is agreed per engagement.
- No client names, testimonials, endorsements or awards.
- No certifications or regulatory status; data-handling controls are agreed per project.
- No team size, and no established workforce across every medical specialty: in-house radiology is the core, and other clinical work is agreed case by case.
- Platform logos are the experience of individual team members, not partnerships.
- Research contributions are credited to the individuals named, not to the lab.

## Research contributions by our clinicians

- **Radiology’s Last Exam.** RadLE 1.0 and 2.0 compare AI image interpretation with radiologists. The second version also examines confidence, reliability and when to defer to a specialist. CRASH Lab (https://crashlab.in/), 2025–2026. CRASH Lab credits Dr Yash Jakhotia as a contributor to RadLE 2.0. RadLE 1.0 paper: https://arxiv.org/abs/2509.25559; RadLE 2.0 report: https://crashlab.in/radle-technicalreport.
- **CheXthought.** A chest X-ray research dataset connecting clinical reasoning with visual attention. Preprint, 2026, arXiv:2604.26288. Research context: Stanford AIMI and collaborators.. Dr Yash Jakhotia and other radiologists in our team contributed as individuals through the Global Radiology Consortium. Read the preprint on arXiv: https://arxiv.org/abs/2604.26288.

## Contact

Camry Labs · contact@camrylabs.ai · https://camrylabs.ai/

Privacy notice for website enquiries: https://camrylabs.ai/privacy