For scientists and engineers who use Julia, or are evaluating it, to combine differential-equation models with neural networks, train them on CPUs and GPUs, and call LLMs from scientific code.
The curriculum is being prepared and reviewed. Enrollment is not open yet.
About this course
Scientific machine learning combines mechanistic models, usually differential equations, with machine-learned components. The result can be more data-efficient, more extrapolable and more interpretable than either alone. Julia's SciML ecosystem is one of the most complete platforms for this work, because its solvers, symbolic modeling tools and automatic differentiation systems are designed to compose.
This course teaches that ecosystem in practice. You will model systems with OrdinaryDiffEq and ModelingToolkit, add neural components with Lux to build universal differential equations and neural ODEs, compute gradients through solvers with SciMLSensitivity using Enzyme, Zygote or Mooncake, and fit them with Optimization.jl. You will try physics-informed neural networks with NeuralPDE, learn interpretable equations back from trained models with DataDrivenDiffEq and SymbolicRegression, and build fast surrogates for expensive simulations. You will use threads, distributed processes and GPUs (CUDA.jl, AMDGPU.jl, KernelAbstractions.jl and Reactant.jl) efficiently, and call LLMs from Julia with PromptingTools.jl and local models through Ollama or LlamaCpp.jl. Finally you will make research code reliable with ReTestItems, Aqua and JET, and package it with Pkg and PackageCompiler or JuliaC. Versions are as of October 2026 (Julia 1.13, ModelingToolkit v11).
What you will be able to do
Formulate a mechanistic model in OrdinaryDiffEq or ModelingToolkit and identify which terms are candidates for a learned component
Build and train a universal differential equation or neural ODE with Lux, SciMLSensitivity and Optimization.jl
Choose an automatic differentiation backend and sensitivity algorithm for a given model size and stiffness, and diagnose gradient failures
Recover an interpretable equation from a trained hybrid model with sparse regression or symbolic regression
Train and run models efficiently on multithreaded CPUs and NVIDIA or AMD GPUs, avoiding type instability and excess allocation
Call hosted and local LLMs from Julia with PromptingTools.jl for structured extraction and code assistance in scientific workflows
Test, benchmark and package a SciML project so that its results are reproducible
Syllabus
6 modules · 20 lessons · about 16 hours
1
Module 1: Julia and SciML Foundations
The language features and modeling tools that SciML builds on.3 lessons
Julia for scientists: multiple dispatch, type stability and environments
Solving ODEs with OrdinaryDiffEq: problems, solvers, stiffness and tolerances
Adding neural networks to mechanistic models and training them.4 lessons
Neural networks in Lux: explicit parameters and states
Neural ODEs and universal differential equations
Gradients through solvers: SciMLSensitivity with Enzyme, Zygote and Mooncake
Training strategies: multiple shooting, optimizer choice and Optimization.jl
3
Module 3: Physics-informed Learning and Discovery
Using physics as a constraint and recovering equations from data.3 lessons
Physics-informed neural networks with NeuralPDE
Equation discovery with DataDrivenDiffEq and SymbolicRegression.jl
Surrogate models for expensive simulations with Surrogates.jl and neural operators
4
Module 4: Performance, Concurrency and GPUs
Making scientific ML fast on the hardware you have.4 lessons
Profiling and type stability: finding allocations and dynamic dispatch
Multithreading with OhMyThreads and distributed ensembles
GPU computing with CUDA.jl, AMDGPU.jl and KernelAbstractions
Compiling models with Reactant for accelerators
5
Module 5: LLMs in Scientific Julia Code
Practical language-model use around research code.3 lessons
PromptingTools.jl: hosted and local models, templates and structured extraction
Local inference with Ollama and LlamaCpp.jl for sensitive research data
AI coding assistants for Julia: what they get wrong and how to check them
6
Module 6: Reliable and Reproducible Research Software
Testing, packaging and sharing SciML work.3 lessons
Testing with ReTestItems, Aqua and JET
Reproducible environments, benchmarks and validation against known solutions
Packaging and deployment: Pkg, PackageCompiler and JuliaC
Exam and certificate
The exam unlocks once you have completed every lesson. It draws 20 questions from a larger bank, spread across the modules, with the answer options shuffled for each attempt. You have 40 minutes and need 80% to pass. After a failed attempt you can try again after 24 hours, up to 3 attempts in total. Feedback is given per module, never as the answer key.
Passing earns a certificate of completion and assessment from the Wirewalk Institute of AI with a unique ID that anyone can check at institute.wirewalk.com/verify. It is not an accredited qualification or a vendor certification. How certification works.
The Wirewalk Institute of AI is independent. This course teaches publicly documented products and is not affiliated with or endorsed by JuliaLang, SciML (DifferentialEquations.jl, ModelingToolkit.jl, SciMLSensitivity.jl), Lux.jl, Flux.jl, Enzyme.jl, Reactant.jl, NVIDIA (CUDA.jl), AMD (AMDGPU.jl), JuliaHub, OpenAI, Anthropic, or Ollama. Product and company names are trademarks of their respective owners.
A free, one-hour, hands-on start built on one realistic job: sorting a property manager's 1,150 monthly maintenance requests. You get an API key the safe way, make real calls with curl and an official Python SDK, run them over forty work orders, read stop reasons, usage and errors, estimate the monthly cost, and decide whether the script is ready. Exercises run on your own provider account for well under a dollar.
A code-first course on integrating LLMs into Python the way production code needs it: uv projects and secrets, the OpenAI, Anthropic and Google GenAI SDKs side by side, async concurrency and streaming, Pydantic structured outputs and tools, retries, local models, embeddings and RAG, FastAPI and pandas integration, testing and evals with pytest, observability, cost tracking and deployment.
Call OpenAI, Anthropic, Gemini and local Ollama models from Perl with real CPAN modules, add streaming, concurrency and validated JSON output to text pipelines, and use AI coding assistants to modernize legacy Perl behind characterization tests.