Luke Shepard

I use data to tell stories.

Saving real lives with artificial intelligence

08 March 2020

It is time. Time for something new.

Last week I joined Tempus. The name means “time” in Latin. I now work with a team of people who are figuring out how to treat diseases by incorporating information from your genetic code. The company uses machine learning to make predictions that help doctors make better treatment decisions for a variety of diseases, starting with cancer.

In a basic scenario, a cancer patient will work with their oncologist to order a test. Tempus will pull “germline” DNA data from a regular blood sample, and combine that with both DNA and RNA from the tumor sample. That gives information about how the tumor differs from the rest of the cells in the body, and the RNA gives hints about actual behvavior of cancer cells. These are combined with clinical data, such as other lab reports, medicines, and other procedures that the patient has undergone. That information all goes into a bunch of models to produce reports that inform oncologists. The doctor can then review recent research or even ongoing trials to see if there is something that would work better for the specific genotype of the patient.

Check out a sample report to get an idea:

This browser does not support PDFs. Please download the PDF to view it: Download PDF.


This mission is insanely compelling. Every day, engineers at Tempus are working to help real patients; they are fighting for their lives and we are helping them do so. I am ridiculously excited to contribute to that effort.

Taking on a smaller team

After seven years of incredible impact, I finally left eSpark this past fall. I was so thrilled to have had opportunity to build such an awesome product that is now in use in every state in the country. I had the privilege to work with incredible people who were devoted to a mission greater than ourselves. At such a small and growing startup, I got a chance to do a little of everything - hiring and growing data science, engineers, designers, product managers, support, HR, operations, and even a few brilliant teachers that wrote the curriculum in use by schools in every state in the country today.

When I began the search for my next role, I was looking for something that would let me get my hands dirty with new techniques in machine learning, was based in Chicago, and would excite me. Tempus delivered in spades on all three.

  • Machine learning at the core of the business. Too many companies try to shoe-horn new technology where it doesn’t belong. I don’t want to just try to use ML for its own sake - rather, I wanted to find a company where success depends on getting it right so we can stay focused on the product mission. Tempus depends on ML for processing the bioinformatics as well as clinical data - and getting good results is essential to the business.

  • Hands-on role. I loved being CTO - it allowed me to try managing many different parts of the business and to become involved in the strategy as well as execution. However, in my next role I wanted to spend more time as an individual - to work with ML technology in a hands-on way without layers of management. At Tempus, I’m leading a small team in an emerging area of new product development.

  • Location - Tempus is a homegrown Chicago company, with a lab and many of the staff in town. However, there is a rapidly growing presence in New York and Redwood City. The former head of Pathology at Google Brain joined last year and founded the Redwood City office, which now contains incredibly talented engineers, data scientists and medical specialists.

I didn’t study biology in college. My only experience with genetic impact is personal - my daughter’s epilepsy is caused by a rare genetic disorder, and I’ve lost several close family members to cancer. But I am surrounded by smart people and have a huge opportunity to learn rapidly and make a difference that literally saves lives. And of course, we are hiring - if you have interest in discussing the company or team feel free to reach out at