Pumas-AI started two years ago with a mission to modernize and improve the efficiency of predictive health analytics, and we are making rapid progress. Check out some of our success stories. Our products have already impacted key decisions in these unprecedented times and there is a growing demand for real-time healthcare analytics, both in Pharmaceutical Drug Development and Precision healthcare at the bedside.
As a result, we are expanding our core team. If you or anyone you know is interested in building products in the healthcare space, please check out our careers page.
Feel free to reach out and speak to us even otherwise. There will be more positions opening up in the future!
Hello, a Pumas-AI employee here, providing unsolicited additional information.
Pumas is a really nice place to be. There’s plenty of cutting-edge scientific and programming expertise around and we’re all enthusiastic about sharing this with oneanother. We’re also globally remote which has allowed me to move back home to my roots. Our cross-discipline nature (pharma, biology, statistics, ML, Julia, dynamics, data science, image analysis, etc.) means that almost no single person has expertise across all that we do so don’t let that deter you from getting in touch about jobs or internships.
Pumas-AI does way more than this but I, myself, am applying all the SciML that’s come out of the Julia eco-system over the last few years to really revolutionise modelling (and thereby, healthcare) across pharmacology. We’re publishing and presenting at conferences, so if you don’t want to close the door on academia entirely then that’s possible. Many of us are also contributing to open-source during work hours which is nice. It’s all pretty cool. Get in touch if you’re interested .
For those who don’t know what this is in a technical sense, let me give a little bit of a high level overview. PharmaceUtical Modeling And Simulation (Pumas) is modeling and simulation of drug pathways and drug response. Quantitative systems pharmacology (QSP) deals with large systems of ODEs (think up to 10,000 ODEs, or PDE discretizations) for predicting the effects of drug targets on patient populations. Physiologically-based pharmacokinetic models (PBPK) deal with the creation of realistic drug concentration models by explicitly modeling the physiology (up to hundreds of ODEs). Pharmacokinetic/Pharmacodynamic (PK/PD) models are generally simpler using compartmental models of 10’s of ODEs, or often less. At each of these scales, differentiable simulation is employed to solve inverse problems, many times on hierarchical models known as nonlinear mixed effects models (NLME), which mix a statistical maximum likelihood problem with the differential equation models. This is optimized in all sorts of ways, using nonlinear optimization methods, Bayesian estimation, etc.
There are so many levels one can work on in this space. Systems of 10 ODEs need to generate “perfect” assembly to simulate fast and make the inverse problems computationally viable. The large systems of ODEs need automated (GPU) parallelism to scale the next generation of problems. New methods, new HPC tooling, new modeling frameworks (ModelingToolkit, Symbolics, etc.) were generated out of solving problems for this domain.
To get a sense of some of what is going on with Pumas, check out the paper on Pumas.jl (https://www.biorxiv.org/content/10.1101/2020.11.28.402297v1). Note that this is the clinical pharmacology aspect of Pumas, which does not include other aspects like preclinical pharmacology, data science / machine learning, data visualization, interactive development / GUIs, etc. which also are in the domain of Pumas-AI.
That’s an understatement. The Pumas team has won 4 of the top awards in the last 3 years at the top pharmacometrics conference. So potential employees and interns take note, I’d call Pumas a good place to get a resume boost .
.
.