Julia was built for the kind of scientific computing that sits at the intersection of performance, composability, and rigor. This makes JuliaCon the natural home for much of what drives Pumas forward. This year, members of our team are presenting work that spans visualization infrastructure and intelligent experimental design, two areas where the Julia ecosystem is moving fast and where Pumas is actively shaping what comes next. At JuliaCon 2026 in Mainz, we are proud to share advances that make Julia-based scientific workflows more powerful, more interpretable, and more practical for real-world research.

Date: August 10 - 15, 2026

Location: Johannes Gutenberg University Mainz, Germany

Podium

Topic: Staged Programming in Pharmacometrics

Date: August 12, 2026 | 12:15 pm - 12:30 pm

Speaker: Andreas Noack Vice President, Product Development, Pumas AI

This talk will outline the modeling class most popular in pharmacometrics, nonlinear mixed-effects modeling (NLME). Andreas will also give an overview of how historical solutions are used, and explain how Julia is well suited for pharmacometrics. Know more

Topic: Makie.jl Highlights: Raytracing, Compute Graphs and Complex Recipes

Date: August 13, 2026 | 10:30 am - 11:00 am

Speakers:
Julius Krumbiegel Senior Product Engineer, Pumas AI
Simon Danisch Creator of Makie.jl

Julius and Simon will walk through a year of major advances in Makie, Julia's native plotting ecosystem. Highlights include RayMakie, a new GPU-accelerated photorealistic ray tracing backend written entirely in Julia, a fundamental refactor that replaces Observables with compute graphs for synchronized, efficient rendering, and an upcoming complex recipe system designed to close longstanding gaps in how users write composable plotting functions. Know more

Topic: Deep Adaptive Experimental Design for SciML

Date: August 14, 2026 | 10:00 am - 10:30 am

Speakers:
Arno Strouwen Senior Product Engineer, Pumas AI
Sebastian MC Software Engineer at JuliaHub & PhD student at University of Bucharest

Arno will present a Julia-native approach to real-time adaptive experimental design using deep neural network policies trained offline to amortize the cost of Bayesian inference. The approach composes Enzyme.jl, Lux.jl, and Reactant.jl without special adaptation — a demonstration of what makes the Julia SciML stack uniquely suited to this class of problem. On a fed-batch bioreactor benchmark, the learned policy outperforms Bayesian D-optimal static designs with a 99.5% win rate. Know more

Whether you are working in scientific visualization, experimental design, or building on the SciML ecosystem, our team is excited to connect at JuliaCon 2026. Come find us in Mainz, catch the talks, and see what the Julia community and Pumas are building next.