A mole seen through a dermatoscope, breaking into the image patches an AI model analyses

Healthcare and life sciences • Advisory

AI consulting with Dr Magnus Lynch

Independent clinical and scientific advice for health tech startups, AI labs, pharmaceutical companies and hospital systems. I am a consultant dermatologist and an Honorary Reader at King's College London, and I have published research on artificial intelligence in medicine since 2020.

Consultant dermatologist, Guy's Hospital Honorary Reader, King's College London Research on AI in medicine since 2020
CambridgeMedicine, University of Cambridge
OxfordDoctorate (DPhil), University of Oxford
FRCPFellow of the Royal College of Physicians
King'sHonorary Reader (associate professor), King's College London
50Research publications
EditorGuest editor, Dermatologic Surgery

Research published in

Who I work with

Clinical and scientific judgement at the points where healthcare AI most often goes wrong.

Health tech startups

Founders building AI products for patients or clinicians.

  • Clinical input on the product and its claims
  • Advice on evidence before launch or fundraising
  • Fitting the product into clinical care

AI labs

Teams developing general or medical models.

  • Advice on clinical benchmarks and evaluation
  • Where models fail and why it matters to patients
  • Safe behaviour in medical conversations

Pharma and biotech

Research, development and partnership teams.

  • AI strategy in dermatology and skin research
  • Image-based endpoints and digital tools
  • Expert review of AI partners and technologies

Hospital systems

Leaders deciding which AI tools to adopt and how.

  • Independent assessment of AI tools
  • Where AI can help clinicians and patients
  • Safe introduction and governance

What I do

Advisory work that gives your team a clear, independent clinical and scientific view.

Advisory board and clinical adviser roles

Ongoing clinical and scientific advice to founders, boards and research teams as a product or programme develops.

You get: regular access to senior clinical and academic judgement.

Evaluation strategy

Advice on how an AI system should be tested: which cases, which comparisons and which outcomes matter to patients, so that the evidence stands up to clinicians, buyers and reviewers.

You get: a written evaluation plan or a review of your existing evidence.

Clinical safety and failure modes

Identifying where an AI system could mislead clinicians or harm patients, drawing on my research into the vulnerabilities of deep learning for skin cancer diagnosis and the ethics of AI diagnosis.

You get: a clear account of the risks and how to reduce them.

Product and clinical workflow advice

Where an AI product fits into a real clinic, who acts on its output, what happens when it is wrong, and what clinicians and patients will trust.

You get: practical recommendations for the product and its clinical pathway.

AI strategy for hospitals and clinics

An independent view of which AI tools are worth adopting, the evidence behind vendors' claims, and how to introduce and monitor them safely.

You get: an assessment and a practical plan.

Expert review

Independent clinical and scientific review of an AI product, study or company, for partnerships, procurement or investment decisions.

You get: a written opinion you can act on.

Publications

Research on AI in medicine

I have written on artificial intelligence in medicine since 2020, from how deep learning works and where it fails in skin cancer diagnosis, to how large language models perform on specialist examinations and the ethics of AI diagnosis. My research group at King's College London also develops and tests its own skin cancer models.

  1. 2025

    Beyond the algorithm: Ethical Challenges in AI-Driven Skin Cancer Diagnosis

    British Journal of DermatologySole author

  2. 2024

    Generative Pretrained Transformer-4, an artificial intelligence text predictive model, has a high capability for passing novel written radiology exam questions

    International Journal of Computer Assisted Radiology and Surgery

  3. 2023

    Systematic review of deep learning image analyses for the diagnosis and monitoring of skin disease

    npj Digital Medicine

  4. 2023

    GPT-4, an artificial intelligence large language model, exhibits high levels of accuracy on dermatology specialty certificate exam questions

    medRxiv (preprint)Senior author

  5. 2022

    Gradient boosting approaches can outperform logistic regression for risk prediction in cutaneous allergy

    Contact DermatitisSenior author

  6. 2021

    Clinically-relevant vulnerabilities of deep machine learning systems for skin cancer diagnosis

    Journal of Investigative DermatologySenior author

  7. 2020

    What is AI? Applications of artificial intelligence to dermatology

    British Journal of DermatologySenior author

Dermoscopic images of skin lesions like the sample images in the browser demonstration

Try it yourself

Try a skin cancer AI in your browser

My research group has put one of its skin lesion classifiers online so that anyone can see how this kind of AI behaves. It is a fine-tuned Inception v3 neural network, trained on 68,011 dermoscopic and clinical images from the ISIC Archive to recognise nine diagnostic groups.

The model downloads once and then runs entirely on your own device: your image is never uploaded. Choose a photograph of a single lesion or try one of the sample images, and the page shows how the network scores it.

  • Runs privately in the browser, on your graphics card or processor
  • Sample images of biopsy-proven lesions to explore
  • Full details of the training, test results and limitations

It is a research demonstration, not a medical device, and it will miss skin cancers.

The same network is the baseline in my article A 10-year-old CNN outperforms frontier LLMs for skin cancer diagnosis, where it beat two frontier language models on 200 biopsy-proven lesions. The article went viral on Hacker News, with 670 points and 274 comments.

Why a clinician-scientist

Advice that combines clinical experience, academic rigour and technical understanding.

Clinical judgement

Over twenty years in medicine, a consultant at St John's Institute of Dermatology, Guy's Hospital, and a busy private practice. I know which errors harm patients.

Academic rigour

An Oxford doctorate, 50 publications and a research group at King's College London. I read evidence the way a journal reviewer does.

Technical understanding

My research group trains and evaluates its own models, so I can engage in detail with your data scientists and engineers.

Clear communication

Years of explaining medicine on television and in the national press. I can explain an AI system's strengths and limits to a board, a clinical team or a patient.

From the clinic

Grounded in a busy clinical practice

I see patients every week at Guy's Hospital and in my private practice in Central London. That gives me a direct view of where AI can save clinicians time and help patients, and where it adds risk: errors that are hard to spot, extra steps in a crowded clinic, unclear responsibility when a tool is wrong, and the trust of patients.

My advice is about how AI works in real clinical care, not only how accurate it is on a test set.

  • The benefits and trade-offs of integrating AI into clinical workflows
  • Patient safety, consent and trust
  • Responsibility and governance when AI is part of care
Dr Magnus Lynch with a patient in clinic
Dr Magnus Lynch in a white coat

How we would work together

  1. Introductory call

    We talk through what you are trying to achieve and whether I am the right person to help.

  2. Written proposal

    I set out the scope, what you will receive, the timeline and the fee.

  3. Advice

    Regular sessions or a defined review, depending on what you need.

  4. Written recommendations

    A clear written summary of my findings and advice.

Questions

What organisations most often ask before we start.

Do you work with organisations outside the UK?

Yes. I am based in London and most of the work is done remotely, so I can work with organisations anywhere.

Will you sign a confidentiality agreement?

Yes. I am happy to sign a confidentiality agreement before you share any details of your product or data.

Do you build or code the AI yourself?

My role is advisory. My research group develops and tests AI models, which keeps my advice technically grounded, and I work alongside your own data scientists and engineers.

How do you fit consulting around clinical work?

I continue to see patients at Guy's Hospital and in private practice, which keeps my advice grounded in current clinical care. I take on a limited number of projects at a time so that each gets proper attention.

Is your expertise limited to dermatology?

My clinical specialty is dermatology, where image-based AI is most advanced. The principles of clinical evaluation, safety and workflow apply across medicine, and I will say plainly when a question needs another specialist.

What does it cost?

It depends on the scope. After an introductory call I send a written proposal with the fee.

Discuss a project

Tell me briefly what you are working on. It helps to include:

  • Your organisation and your role
  • The product, model or decision involved
  • What you would like help with, and by when

Email contact@drmagnuslynch.com

I reply personally to every enquiry, usually within two working days.

Dr Magnus Lynch MA(Cantab) DPhil(Oxon) MRCS FRCP, Consultant Dermatologist and Honorary Reader at King's College London. See about Dr Lynch, research and publications.

Discuss a project