Clinical documentation

made simple.

Almanac Chat is a research collaboration between like-minded researchers to bring robust and transparent generative AI to clinical medicine.



State of the art

Built from the ground up

We recognize that the integration of AI language models in healthcare has the potential to revolutionize patient care, research, and medical education. However, it is crucial to proceed with caution and rigorously evaluate these systems to address concerns regarding accuracy, bias, privacy, and the potential for unintended consequences.

That's why we're building Almanac Chat - a multi-institutional effort to better understand the potential and limitations of these multimodal language models in healthcare. Our goal is to comprehensively evaluate these systems from the workbench to the bedside, in order to ensure that they are safe, effective, and equitable.

7 billion +
Clinical Reports
1.5 million +
80K +
Medical articles
18K +

Fully transparent

Meaningful and Reproducible Benchmarks

The current landscape of generative AI in clinical medicine is fragmented and siloed. Despite the promise of these technologies, there exists a lack of transparency and reproducibility in the field.

To address these challenges, we're developing Almanac Chat in the open, with a focus on three core principles:

Safety and Alignment

We recognize the powerful role that large language models can play in the clinic, as well as the potential dangers of careless deployment and use. As physicians it remains our responsibility to ensure that these technologies are made safe and effective for our patients.

Reproducibility and Collaboration

We believe that real progress is made when people work together in a continuous and iterative process, towards a common goal. As such, we aim to develop and establish a suite of baselines and meaningful benchmarks to encourage open and reproducible research for the benefit of the entire medical community.

Accessibility and Cost

Large language model research can be prohibitively expensive and inaccessible to many healthcare professionals and researchers. We aim to democratize access to these technologies by developing models that can be trained, deployed, and evaluated on consumer hardware.

Built collaboratively

Our Researchers

We're a small team of physicians and engineers dedicated to improving and evaluating the potentials and pitfalls of generative AI in medicine.

  • Cyril Zakka, MD

    Stanford Medicine

  • Akash Chaurasia, BS

    Stanford Engineering

  • Rohan Shad, MD

    Penn Medicine

  • Michael Moor, MD PhD

    Stanford Engineering

  • Katie Link, BS

    Hugging Face

Advised by

  • William Hiesinger, MD

    Stanford Medicine

  • Pranav Rajpurkar, PhD

    Havard Medical School

  • Jure Leskovec, PhD

    Stanford Engineering