JuMP-dev 2026
JuMP-dev 2026 will be held Sunday, May 31 and Monday June 1, 2026 in Edinburgh, Scotland.
This is the weekend prior to SIAM OP26, which will be held June 2-5, 2026 in Edinburgh, Scotland.
JuMP-dev will be co-located with the 2026 HiGHS Workshop.
Outline
The purpose of JuMP-dev is to bring together students, researchers, and practitioners with interests in the methodological, algorithmic, and software aspects of JuMP and related packages. In particular, we invite new contributors and those who have not met the core development team.
All participants will uphold the JuMP Code of Conduct.
How do I attend?
To attend JuMP-dev, you must purchase a ticket.
If you use JuMP in a professional capacity, please consider purchasing a
“Supporter ticket”. If you purchase a supporter ticket, please email your
company logo to jump-dev-2026@googlegroups.com, and we will include it in our
promotional materials.
To request a visa invitation, please write to jump-dev-2026@googlegroups.com
with the subject line Visa invitation letter.
Talks
Talk submissions are now closed.
Schedule
The workshop is structured as two days of talks. Unlike previous years, there will not be a hackathon; the JuMP core contributors will be attending SIAM OP26.
- Saturday 30 May: pre-conference excursion
- Sunday 31 May: talks at the John McIntyre Conference Centre, dinner at the Salisbury Arms
- Monday 1 June: talks at the John McIntyre Conference Centre
On Monday 1 June, the 2026 HiGHS Workshop will be taking place in parallel in an adjacent room. Participants are welcome to move between streams.
Talk schedule
The full set of recorded talks are available as a YouTube playlist.
| Sunday 31 (Prestonfield) | Monday 1 (Prestonfield) | Monday 1 (Holyrood) | |
|---|---|---|---|
| 08:30 | Coffee |
Coffee |
Coffee |
| 09:00 | |||
| 09:15 | Introduction and welcome Oscar Dowson |
||
| 09:30 | ElectricityEmissions.jl: Calculating carbon intensity signals on the power grid Noah Rhodes [slides][video] |
DiffOpt.jl: Differentiable optimization for JuMP Joaquim Dias Garcia [slides][video] |
What's new in HiGHS Julian Hall [slides] |
| 09:45 | Optimal Planning for Steel Industry Decarbonization: A Mixed-Integer Linear Programming Approach in Julia Vinícius Justen Pinto [slides][video] |
||
| 10:00 | Optimize Your Fleet with Odoo and JuMP Benoît Legat [slides][video] |
DisjunctiveProgramming.jl’s New Suite of Solution Methods Daniel Nguyen [slides] |
HiPDLP: A PDLP solver for HiGHS Yanyu Zhou [slides] |
| 10:15 | DuckDB as backend to build optimization models in JuMP.jl Abel Siqueira [slides][video] |
DisjunctiveNet.jl: Neural Symbolic Learning via Differentiable Convexified Optimization Layers Can Li [slides] |
|
| 10:30 | Morning tea |
Morning tea |
Morning tea |
| 10:45 | |||
| 11:00 | UnoSolver.jl, a unified SQP/barrier solver for nonlinearly constrained optimization Charlie Vanaret [slides][video] |
Encoding Matters when Embedding NNs in Optimization Problems Chelsea Sidrane |
Tallarium Lucas Biton |
| 11:15 | InfiniteMathOptAI.jl: Embedding Surrogates for Infinite-Dimensional Optimization Joshua Pulsipher [slides] |
||
| 11:30 | Reduced-space nonlinear optimization via the implicit function theorem Robert Parker [video] |
Optimization Problem Surrogates with Graph Transformer Networks and integration within simulation workflows Jose Daniel Lara |
Prescriptive Analytics ought to be in the database Soren Little |
| 11:45 | |||
| 12:00 | Batched optimization solvers: a tour d'horizon François Pacaud [slides][video] |
Toward Extending EAGO.jl for Exploiting GPU Parallelism Dimitri Alston [slides] |
Optimisation Problems at Edinburgh Airport Ivona Gjeroska and Bárbara Cotrim Rodrigues |
| 12:15 | |||
| 12:30 | Lunch (provided on-site) |
Lunch (provided on-site) |
Lunch (provided on-site) |
| 14:00 | Second-Order GPU solver for Burer-Monteiro Benoît Legat [slides] |
Open Energy Benchmark: 2025 results, HiPO performance, and call for benchmarks Daniele Lerede [slides] |
|
| 14:15 | Antipode JuMPing - How Norwegian authorities can use JADE.jl for better decision making Jarand Hole [slides][video] |
GraphsOfConvexSets.jl: Optimisation over Graphs of Convex Sets with JuMP Charles van Hees [slides] |
|
| 14:30 | Automatic Decomposition of JuMP Models using TimeStructDecomposition.jl Truls Flatberg [slides][video] |
First-order Convex (mixed-integer) optimization: FrankWolfe.jl and Boscia.jl Deborah Hendrych [slides] |
Recent developments and future plans for HiPO Filippo Zanetti [slides] |
| 14:45 | Advanced Model Composition in ApplicationDrivenLearning.jl Giovanni Amorim [slides][video] |
TenSolver.jl: A tensor network based QUBO solver with JuMP integration Iago Leal de Freitas [slides] |
|
| 15:00 | Modelling and Solving Mathematical Programs with Complementarity Constraints. Anton Edvinovich Pozharskiy [slides][video] |
PEPit.jl: Computer-Assisted Worst-Case Analysis of First-Order Optimization Algorithms in Julia Shuvomoy Das Gupta [slides] |
A parallel MIP solver for HiGHS! Mark Turner [slides] |
| 15:15 | Adding array support for JuMP’s Automatic Differentiation Sophie Lequeu [slides][video] |
||
| 15:30 | Afternoon tea |
Afternoon tea |
Afternoon tea |
| 15:45 | |||
| 16:00 | Can ExaModels Power JuMP on GPUs? Sungho Shin [video] |
Local heuristics for the HiGHS MIP solver Ben Champion [slides] |
|
| 16:15 | JuMP's macros: why are they needed? Oscar Dowson [slides][video] |
The 2026 JuMP-dev prize The JuMP developers |
|
| 16:30 | |||
| 16:45 | |||
| 17:00 | Dinner |
Workshop ends |
Workshop ends |
SIAM OP26
We also organized two sessions at SIAM OP26, which took place in the week following JuMP-dev. There were six talks, but no video recordings:
- The state of JuMP, by Miles Lubin [slides]
- MathOptAI.jl: Embed Machine Learning Predictors into a JuMP Model, by Oscar Dowson [slides]
- InfiniteMathOptAI.jl: Embedding Surrogates for Infinite-Dimensional Optimization, by Joshua Pulsipher [slides]
- ApplicationDrivenLearning.jl a Framework to Train Forecast Models with Application-driven Cost Functions, by Giovanni Amorim, Joaquim Dias Garcia, and Alexandre Street [slides]
- The State of ExaModels, by Sungho Shin, Alexis Montoison, and François Pacaud [slides]
- Adding array support for JuMP’s Automatic Differentiation, by Benoît Legat [slides]
Venue
The workshop will take place in downtown Edinburgh.
Talks
JuMP-dev will be held at the John McIntyre Conference Centre. There are a wide range of accommodation options within walking distance.
Use the entrance on the south-west corner of the building. The registration desk is up the stairs. The lecture theatres are straight ahead, through a few sets of doors. (Go past the terrace on your right.) From the lounge area where the food will be served, the JuMP talks will be in the Prestonfield room on right and the HiGHS talks will be in the Holyrood room on your left.
Pre-conference excursion
On Saturday May 30, we will walk up Arthur’s Seat for a view of the city.
Meet at the Holyrood Carpark at 2pm.
Dinner
On Sunday night we will meet for an informal dinner at the Salisbury Arms. The pub is a short walk from the conference venue.
Accommodation
There are a wide variety of accommodation options nearby.
Accommodation is available at the conference venue, but it is open only from Monday 1 June. (You will need to book somewhere else for the weekend.) For more details, see Summer stays at the University of Edinburgh.
Program committee
- Oscar Dowson (Program chair)
- Mathieu Besançon (Inria)
- Joaquim Dias Garcia (Soma Energy)
- Benoît Legat (UCLouvain)
- François Pacaud (Mines Paris-PSL)
- Chelsea Sidrane (KTH)
- Julian Hall (University of Edinburgh; local organization)
- Filippo Zanetti (University of Edinburgh; local organization)
- Ivet Galabova (University of Edinburgh; local organization)
To contact the program committee, write to jump-dev-2026@googlegroups.com.
Sponsors
