Julia Dispatch

In this episode of the Julia Dispatch, hosts Christopher Rackauckas and Michael Tiemann dive into the world of machine learning and compiler engineering with core developers Billy Moses and Avik Pal.We pull back the curtain on the future of high-performance computing in Julia, tracking the evolution of revolutionary compiler tools like Enzyme.jl and Polygeist into their latest groundbreaking collaboration: Reactant.jl. Billy shares how his journey from writing quick-and-dirty Python-to-Java source rewriters in high school eventually led to a PhD at the MIT Julia Lab and a professorship at UIUC. Avik recounts how his background in Google Summer of Code and scientific machine learning (SciML) exposed the structural fragility of language-level neural network optimization—and how Reactant stepped in to solve it.The group explores how Reactant breaks the traditional boundaries of Domain-Specific Languages (DSLs) to automatically optimize generic, loop-heavy, and mutating Julia code directly into MLIR subgraphs, paving the way for next-generation hardware acceleration.Hosts: Chris Rackauckas & Michael TiemannEditor: StaziOfficial Website: https://juliadispatch.fmGitHub Repository: https://github.com/JuliaDispatch/YouTube: https://www.youtube.com/@JuliaDispatchListen via RSS: https://anchor.fm/s/fc63539c/podcast/rss#JuliaLang #MachineLearning #Compilers #SciML #Reactant #Enzyme #OpenSource #TechPodcast

Happy New Year! In this episode, we welcome back Leticia Madureira, a PhD student at Carnegie Mellon University, to explore her journey into quantum chemistry, her passion for Julia, and her work in computational chemistry. Leticia shares her experiences as a PhD student, her contributions to the Julia ecosystem, and her vision for the future of quantum chemistry simulations.

Leticia Madureira is a PhD candidate in Chemistry at Carnegie Mellon University, specializing in computational quantum chemistry. Her work focuses on electronic structure methods to study chemical processes like polymerization and photodegradation. She collaborates with the Julia Lab to develop open-source, high-performance tools for quantum chemistry, including packages like BasisSets.jl and OohataHuzinaga.jl (under the HartreeFoca organization).Leticia aims to make computational chemistry accessible through teaching and open-source development, believing that education empowers scientists and fosters innovation in the field.

Recorded on: 2025/10/10

Hosts: Chris Rackauckas, Michael Tiemann

Editor: Stazi

Find us everywhere:

https://juliadispatch.fm

https://github.com/JuliaDispatch/

https://www.youtube.com/@JuliaDispatch

https://anchor.fm/s/fc63539c/podcast/rss

Merry Chrismas to all who observe! Today, we dive into the world of Makie.jl with its core developers, Simon Danisch and Julius Krumbiegel. We explore the origins of Makie, its unique design philosophy, and its role in shaping the future of visualization in Julia. Simon and Julius share their journeys into Julia, the challenges of building a flexible and performant visualization framework, and their vision for Makie’s future—including ray tracing, higher-level APIs, and integration with the broader Julia ecosystem.

Simon Danisch began his studies in Cognitive Science at the University of Osnabrück in 2010, specializing in computer vision and machine learning. In 2012, he discovered Julia as a language that combined high performance for interactive computing, seamless GPU integration, and an elegant alternative to object-oriented programming for mathematical applications. Julia has been his primary language ever since. While working with C++ on machine learning projects in 2011, Simon identified a need for interactive, user-friendly tools for data visualization and model parameter manipulation. This inspired his Bachelor’s thesis, where he developed an early predecessor to Makie.jl—some of which remains in use today. Since then, he has made significant contributions to Julia’s graphics, GPU, and plotting infrastructure, and is the author of numerous related packages.

Julius Krumbiegel joined the Makie project in 2019, where he developed its layout system (GridLayoutBase.jl) and many of the GUI objects originally part of the standalone MakieLayout.jl package. With a background in psychology and vision science, his work focuses on 2D plots, visual quality, and usability, including Makie’s default themes, plot recipes, text and figure layouting, and vector graphics output via CairoMakie.jl. He is also the author of several widely used Julia packages, including Chain.jl, DataFrameMacros.jl, ReadableRegex.jl, Animations.jl, and SankeyMakie.jl.

Makie's website: https://makie.org/website/

Recorded on: 2025/11/25

Hosts: Chris Rackauckas, Michael Tiemann

Editor: Stazi

Find us everywhere:

https://juliadispatch.fm

https://github.com/JuliaDispatch/

https://www.youtube.com/@JuliaDispatch

https://anchor.fm/s/fc63539c/podcast/rss

Join Chris Rackauckas and Stefan Kastanov for an in-depth discussion of recent developments in the Julia programming language community, with a special focus on JuliaCon 2025 and exciting technical advances.

🎤 JuliaCon 2025 Recap

• The unique grassroots nature of JuliaCon and why it's different from typical academic conferences
• How the community-driven approach fosters collaboration over competition
• The impact of geopolitical issues on attendance and the rise of JuliaCon Local events
• JuliaCon Local Paris registration now open!

💻 Technical Deep Dives

• State of Julia talk: The famous "slide deck karaoke" featuring 20+ core contributors
• Cody's work on trim and compilation tooling for deploying Julia to real devices
• Sparse arrays developments and ecosystem highlights
• Fixed-size arrays and why they represent the future of Julia's array interface

🤖 AI-Powered Development in Julia

• Chris shares his extensive experience using Claude for repository maintenance across 200+ packages
• Best practices for sandboxing AI tools (including the "justme_claude" GitHub account approach)
• How to safely incorporate AI into your Julia development workflow
• Setting up effective agent instructions for code maintenance

🔬 Automatic Differentiation Updates

• Mooncake.jl emerging as a strong contender to replace Zygote
• Support for mutation in AD while maintaining high-level Julia code
• Comparison of different AD approaches in the Julia ecosystem

🎮 Game Development in Julia

• Entity-Component System (ECS) frameworks coming to Julia
• Rex game engine development
• Commercial game development with Julia on Steam

📝 This Month in Julia Newsletter: https://discourse.julialang.org/t/this-month-in-julia-world-2025-06-07/131412

🎟️ JuliaCon Local Paris Registration: [Link to be added]

🔧 Mentioned Tools & Packages:

• Mooncake.jl - Modern automatic differentiation
• SandboxJL - Safe sandboxing for AI tools
• Rex - Game engine in Julia
• Fixed-size arrays

00:00 - Introduction & Welcome00:33 - JuliaCon 2025 Overview - Grassroots Community Nature01:34 - Quality of JuliaCon Talks vs Academic Conferences02:18 - Community Collaboration & Interoperability03:11 - Growth of JuliaCon Local Events & Visa Issues04:03 - JuliaCon Local Paris Registration Open04:25 - JuliaCon Talk Highlights Discussion04:46 - Focus on Language, Compiler & Type System Topics05:36 - State of Julia Presentation - Slide Deck Karaoke06:13 - Behind the Scenes of State of Julia Creation07:26 - Cody's Work on Trim & Compilation Tooling39:53 - AI Tools in Julia Development - Claude for Repository Maintenance40:29 - Setting Up Agent Files & Instructions41:47 - Sandboxing AI Tools - The "justme_claude" GitHub Account43:01 - Safe Practices for Using AI in Development43:42 - Entity-Component Systems (ECS) Framework Discussion44:23 - Game Development in Julia44:51 - Commercial Game on Steam with Julia46:44 - Mooncake.jl for Automatic Differentiation46:48 - Wrap Up & Future Episodes47:22 - FerriteCon Copenhagen & Contributing to Newsletter48:02 - Closing Remarks

Chris Rackauckas is a Research Affiliate and Co-PI of the Julia Lab at MIT, VP of Modeling and Simulation at JuliaHub, and Lead Developer of the SciML Open Source Software Organization.

Stefan Kastanov is a Julia community contributor and regular co-host of Julia Dispatch.

🌐 Julia Language: https://julialang.org💬 Julia Discourse: https://discourse.julialang.org🐙 JuliaHub: https://juliahub.com🔬 SciML: https://sciml.ai

📚 New to Julia? Check out our beginner-friendly episodes and tutorials on the channel!

👍 If you enjoyed this episode, please like, subscribe, and share with fellow Julia enthusiasts!

#Julia #JuliaLang #JuliaCon #Programming #ScientificComputing #OpenSource #MachineLearning #AutomaticDifferentiation #GameDev

Episode HighlightsLinks & ResourcesTimestampsAbout the HostsConnect With Us

Today we're joined by Jesper and Philip from Cognibotics, a Swedish robotics company that has (co-)developed something truly unique: Juliet, a statically-typed dialect of Julia designed specifically for real-time robot programming, along with its runtime system Romeo. Our guests share their personal journeys to Julia and explain how they've adapted the language for industrial robotics applications. They introduce us to Juliet's fascinating features, including static typing, real-time garbage collection, and perhaps most impressively, reversible robot programming that allows forward and backward program stepping for interactive debugging. The conversation explores how they use multiple dispatch for hardware abstraction across different robot backends and hardware targets, their partnership with Chinese robot manufacturer Estun for real-world deployment, and their approach to real-time systems programming. Jesper and Philip also discuss the technical challenges they've overcome, including optimization constraints and the limitations they've had to accept when adapting multiple dispatch for real-time environments. They share their plans for open source access, educational applications including course development at Lund University, and how their work bridges the gap between academic research and industrial automation.

Jesper and Philip are developers at Cognibotics, a Swedish robotics company focused on advanced robot programming solutions. Their work centers around Juliet, a statically-typed dialect of Julia they've created specifically for real-time robot programming, paired with the Romeo runtime system. Cognibotics has formed partnerships with industrial robot manufacturers including Estun, bringing their innovative programming approach to real-world manufacturing applications. The company is also developing educational programs, including plans for courses at Lund University and beta program access for researchers and developers interested in real-time Julia applications.

Cognibotics: https://www.cognibotics.com/en

Juliet & Romeo: https://www.cognibotics.com/en/products/juliet-and-romeo

Lund University: https://www.lunduniversity.lu.se/

Recorded on: 2025/06/06

Hosts: Chris Rackauckas, Michael Tiemann

Editor: Stazi

Find us everywhere:

https://juliadispatch.fm

https://github.com/JuliaDispatch/

https://www.youtube.com/@JuliaDispatch

https://anchor.fm/s/fc63539c/podcast/rss

In this episode, Chris Rackauckas and Michael Tiemann welcome back Firoozeh Dastur, a theoretical mathematician and PhD candidate from Germany working on tropical geometry. This marks the first time someone has returned to the Julia Dispatch podcast! Firoozeh shares her journey from Pakistan to Germany, her pivot from K3 surfaces to tropical geometry, and how she discovered Julia during the pandemic. She explains what tropical geometry is, her early struggles with Julia, and how the Julia Slack community became crucial to her learning process. The conversation then shifts to her work on computational algebra and HPC workflows, where she discusses the frustrations with existing HPC tools that led her to develop DistributedWorkflows.jl. She introduces us to Petri nets as an alternative to DAGs for managing cyclic workflows and explains how her package aims to make HPC accessible to everyone. Firoozeh also shares insights about overcoming perfectionism in open source development, the importance of releasing early (inspired by Airbnb's launch story), and her thoughts on differential equations as Turing complete systems.

Firoozeh Dastur is a theoretical mathematician specializing in algebraic and tropical geometry with a strong focus on leveraging computer algebra systems for advanced mathematical research. Her other interests include high-performance computing (HPC) tools, with a particular goal of making distributed computing accessible to all domain scientists. As a passionate educator, Firoozeh is committed to fostering an inclusive learning environment that empowers the next generation of researchers and innovators. In her spare time, she develops Julia libraries that reflect her interests in distributed computing and computer algebra systems.

DistributedWorkflows.jl: https://github.com/JuliaServices/DistributedWorkflows.jl

Julia Slack community: https://julialang.org/slack/

NEMO.jl: https://github.com/Nemocas/Nemo.jl

Oscar.jl: https://github.com/oscar-system/Oscar.jl

SPACK: https://spack.io/

Recorded on: 2025/05/20

Hosts: Chris Rackauckas, Michael Tiemann

Editor: Stazi

Find us everywhere:

https://juliadispatch.fm

https://github.com/JuliaDispatch/

https://www.youtube.com/@JuliaDispatch

https://anchor.fm/s/fc63539c/podcast/rss

Chris and Guillaume discuss the contents of the "This month in Julia world - 2025-08". We cover the upcoming JuliaCon Local Paris event, Julia 1.12 beta release updates, major compiler improvements with JuliaLowering.jl, performance optimization discoveries, and ecosystem growth with new packages like TestPicker.jl, BorrowChecker.jl, and AI integration tools. We also dive deep into technical discussions about optional modules, static arrays, and the ongoing focus on reducing latency across the Julia ecosystem.

This month in Julia world - 2025-08: https://discourse.julialang.org/t/this-month-in-julia-world-2025-04/128859

Recorded on: 2025/05/13

Hosts: Chris Rackauckas, Guillaume Dalle

Editor: Stazi

Find us everywhere:

https://juliadispatch.fm

https://github.com/JuliaDispatch/

https://www.youtube.com/@JuliaDispatch

https://anchor.fm/s/fc63539c/podcast/rss

Good morning, good evening, good night, wherever you are in the world! Today we welcome Jacob Zelko, who takes us on a fascinating journey from his early days as a biomedical engineering student through his work at the CDC during the COVID-19 pandemic, and into his current exploration of applied category theory. Jacob shares how his pandemic response work led him to seek more fundamental approaches to public health problems, ultimately discovering category theory as a potential framework for composing different models and analyses. We also dive into his role as leader of the Julia Health Organization, his experience as a Twitch streamer known as "The Cedar Prince," and his vision for using Julia to bridge the gap between high-performance computing and accessible health research tools.

Jacob Scott Zelko is a MS student in Applied Mathematics at Northeastern University (NEU) and a trainee of NEU's Roux Institute. Prior to this, he has worked for Georgia Tech Research Institute as a Health Data Analytics and Informatics Researcher and as a Research Engineer for the Centers for Disease Control Office of Science. In the Julia community, he is the leader of the JuliaHealth organization, a community of Julia developers who use the Julia programming language to improve medicine, health care, public health, and biomedical research. Additionally, he is the co-admin for the Julia Language's GSoC program and is the Community Manager for the AlgebraicJulia ecosystem.

Julia Health Organization: https://juliahealth.org/

Jacob's personal website: https://jacobzelko.com/

Jacob's GitHub: https://github.com/TheCedarPrince

Recorded on: 2025/05/07

Hosts: Chris Rackauckas, Michael Tiemann

Editor: Stazi

Find us everywhere:

https://juliadispatch.fm

https://github.com/JuliaDispatch/

https://www.youtube.com/@JuliaDispatch

https://anchor.fm/s/fc63539c/podcast/rss

Join Chris Rackauckas and Stefan Krastanov as they explore the latest developments in the Julia programming language ecosystem from the March 2025 "This Month in Julia World" newsletter.Link to the newsletter: https://discourse.julialang.org/t/this-month-in-julia-world-2025-03/127821🔥 Key Topics Covered:Julia 1.12 Beta 1: Major features including experimental --trim for smaller binaries, stackless compiler improvements, and new threading defaultsBLAS Lazy Loading: Groundbreaking changes to reduce startup time and memory consumption by loading linear algebra libraries on-demandREPL Improvements: Complete overhaul of auto-completion using the new Julia syntax parser, fixing numerous long-standing bugsPerformance Enhancements: Invalidation reduction efforts, faster code coverage, and compiler optimizationsEcosystem Updates: Julia on Google Colab with TPU support, Documenter.jl public keyword filtering, AlgebraOfGraphics.jl v0.10🚀 Highlights:How the new --trim flag enables dramatically smaller Julia binariesWhy BLAS lazy loading is a game-changer for startup performanceThe transition from Lisp-based to pure Julia parsing infrastructureReal-world impact of compiler improvements on large symbolic expressionsJulia running natively on TPUs through Reactant.jl📅 Upcoming Events:JuliaCon Local Paris (October 2025)Quantum Information Science Summer School in Amherst (June 2025)RUST Julia Meetup in EindhovenPerfect for Julia developers, data scientists, and anyone interested in high-performance computing and the latest language developments. Whether you're a beginner or advanced user, you'll find valuable insights into Julia's evolving ecosystem.#Julia #Programming #HighPerformanceComputing #OpenSource #datascience Recorded on: 2025/04/24Hosts: Chris Rackauckas, Stefan KrastanovEditor: StaziFind us everywhere:https://juliadispatch.fmhttps://github.com/JuliaDispatch/https://www.youtube.com/@JuliaDispatchhttps://anchor.fm/s/fc63539c/podcast/rss

Today, we're joined by Stefan Krastanov who is not only the main developer of QuantumClifford.jl, but also the author of the "This month in Julia world" newsletter. Stefan shares with us what passions him about physics, the combination of mathematical concepts and hands-on signal processing and what attracted him to Julia. And he confides in us how the Julia newsletter is a productive form of procrastination, how his own appreciation and understanding grew from being the newsletter editor, and how important it is to improve the communication and documentation to make a technology appealing, in particular to new comers.

Stefan Krastanov is an assistant professor of at the University of Massachusetts Amherst. Before that, he was a postdoc at MIT/Harvard and obtained his PhD in physics at the Yale Quantum Institute. Stefan works on the design, control, and optimization of quantum hardware for computation and networking, from its analog physical description up to the compilation of error-corrected logical circuitry running on it. His research centers around leaky abstraction boundaries between the many layers of technologies making up the field of quantum computing and quantum information science. To this end, he has authored multiple related Julia packages. And Stefan is the author of the "This month in Julia world" newsletter.

Recorded on 2025/04/09

Hosts: Chris Rackauckas, Michael Tiemann

Editor: Stazi

Find us everywhere:

https://juliadispatch.fm

https://github.com/JuliaDispatch/

https://www.youtube.com/@JuliaDispatch

https://anchor.fm/s/fc63539c/podcast/rss