For everybody who uses modelling and analytics to help their organisation and their customers; and everyone interested in the practice of OR.
Room 40, Building 324, stream 46
|
Monday 1st July |
Tuesday 2nd July |
Wednesday 3rd July |
|
|
A 8.30-10.00 |
Generative AI For OR People: Disruptor, Enabler, Or Distraction? |
Not the Excellence in Practice Award: lifting the lid on practice |
|
|
B 10.30-12.00 |
What’s on your mind? – getting what you want from this conference |
Practical Skills: Knowledge Sharing Tutorial -2- Coding; optimisation tips |
|
|
C 12.30-14.00 |
Practical Skills: Knowledge-Sharing Tutorials -1- Streamlit; Kubernetes |
From Academia To Practice: Reflections From Those Who Have Taken The Leap |
|
|
D 14.30-16.00 |
The Quest for the Right OR Professionals – A Hiring Perspective |
‘Making an Impact’ (MAI) is a stream for everybody who uses modelling and analytics to help their organisation and their customers. It is dedicated to helping practitioners improve their effectiveness, develop their skills and knowledge, and do a better job of OR in practice. Every session of the MAI stream has been designed to provide value, whether that is through discussion and reflection on typical practical challenges, tutorials in techniques or software others have found useful, or meeting and sharing experiences with peers.
The MAI sessions and tutorials are described below. Please contact us on
MA-46 Fast And Furious: Lightning Talks
Get the flavour of the whole conference in one 90’ session! Confirmed speakers so far: Laura Albert (Uni of Wisconsin), Paul Bouman, Gökhan Ceyhan (JustEat), Ahmet Çınar (Trendyol), Claudio Gambella (FICO),Caroline Jagtenberg (Vreije Universitat Amsterdam), Mirjam Meijer (YW4OR, Kühne Logistics University), Richard Oberdieck (Circle Banking), Hans Schwartzlose and Kong You Liou (Ørsted), Carolina van Weerd (TNO), Johan Kellerth Fredlund (Jeppesen), Héctor Cancela (Universidad de la República (Uruguay)).
At a conference like EURO, it is impossible to catch up with everything; and all too easy to stick to a single stream within your usual professional specialism. This session showcases some of the breadth of OR in action, with a set of 5-minute presentations handpicked to inspire, interest and extend your knowledge.
Some of the presentations are designed for this session only, others are abbreviated versions of talks appearing elsewhere in the conference so you can catch up with the speaker’s main talk later on, if you wish.
The talks cover a wide range of OR methodology across a variety of practical application areas. Full details: LightningTalks
MB-46 Speed Networking
Meet your peers in a friendly, informal way, through discussions in groups of 2-3 persons with pairings changing after a fixed time span of a few minutes.
OR is a team business, and knowing people you can turn to for ideas, feedback and support makes all the difference. But it is not always easy building your network, especially if you are shy or feel that you are an outsider. This welcoming session is a way of overcoming the barriers to networking, and enjoying yourself while you do it.
If you have business cards, do bring them along; if not, we’ll have plenty that you can use if you choose.
MC-46 Practical Skills: Knowledge-Sharing Tutorials -1-
Expand your skills with these two 45' practical sessions: Building And Sharing Apps: An Introduction To Streamlit, and OR Goes Kubernetes: How To Run Long Running Optimization Jobs In The Cloud
(i) Building Interactive Apps Quickly: An Introduction To Streamlit (Sander van Aken, Flixbus)
Making impact with OR often requires an end-user or other stakeholders to interact with the models and algorithms you build, or with its outcomes. With interactive applications, we can leverage the true power of human-machine collaboration, e.g. by letting users guide your optimization model with their knowledge, or enabling them to explore multiple solutions. Developing fully-fledged front-end applications - or even discovering what is the right thing to develop - can however take up a tremendous amount of time. For some analyses or projects, it is not even worth the effort.
The Python-based framework Streamlit could be your new companion in these endeavours. In this tutorial, we demonstrate how we can use it to quickly develop a first version and iterate on that.
(ii) OR Goes Kubernetes: How To Run Long Running Optimization Jobs In The Cloud (Jonas Witt DHL)
While response time often matters in OR applications, in many cases it does not (think network design, tactical capacity planning, …). In these use cases, model usage will be dispersed over time and not warrant the constant provisioning of a high-end compute resource at all times. In our OR applications at DHL Group, we experienced that this often led to some boilerplate functionality that had to set up across use cases: Sequencing of requests, automatic deployment of workloads, status monitoring, result retrieval, etc. In this tutorial we will share what we are currently building to replace boilerplate code with a scalable, modern tech-stack to support the deployment of tactical decision support tools.
MD-46 The Quest for the Right OR Professionals – A Hiring Perspective
Panel members: Sander van Aken (Flixbus, chair), Henning Blunck (DHL), Michael Lindahl (QAMPO), Susanne Heipcke (FICO), Pol Arias Melià (JustEat Takeaway)
In this panel session, we explore how recruiters of OR professionals, data scientists or analytics people go about it, and address burning questions relevant to both those recruiting professionals, as well as those looking to secure a new industry position.
How do companies and teams come to define the profile to hire, crafting the job ad? What skillset and traits do you look for, and how to you define that? How do you design a hiring process covering those aspects, and what are the challenges and best practices for the recruiting team themselves? What advice would the hiring team give their counterparts?
Join us to learn about each other’s challenges, best-practices, and ask your own questions during a Q&A.
TA-46 Generative AI For OR People: Disruptor, Enabler, Or Distraction?
Large Language Models (LLMs) have taken the world by storm since the release of ChatGPT. These models can work with large amounts of unstructured data, such as text and images. Their output is easy to understand, but they cannot reason and apply logic.
On the other hand, Mathematical Optimization Models require structured and complete data but provide optimal output. However, it can often be a challenge to collect the necessary data in an organization and ensure that users can interact with the final application and understand the output.
Can these very different models complement each other when we build OR applications?
This session will be led by Michael Lindahl. Michael has been building OR-powered decision applications across different industries for more than 10 years. Since the release of ChatGPT, he has been exploring the synergies between simple optimization and LLMs by building the application findgaven.ai on top of their API's.
Michael will give an introduction to large language models, explaining their capabilities, the pitfalls, and how they can fit into the OR toolbox.
This will lead to discussion, with the audience invited to share not only their questions but also their own experiences, aspirations, and concerns.
TB-46 What’s on your mind? – getting what you want from this conference
At a conference like this, we want to interact with, and learn from, each other. Unfortunately, it’s sometimes just not so easy to bump into and connect with the right people to discuss the topics you want to learn about.
In this open space session, it’s the participants who bring in, and discuss about, topics which are on their mind. So you want to learn from others about a particular topic related to making an impact with OR in practice? Join us in this session where you set the agenda. At the start, we will collect the topics which are on your and others’ mind. Using this, we split in smaller groups, and you join the one(s) in which you are interested. As such, it’s you who shape the discussions, and can get a lot out of it.
No topic, just curious what’s on others’ minds? There will already be some kick-starter points available.
TC-46 From Academia To Practice: Reflections From Those Who Have Taken The Leap
Panel members include (TBC): Ruben Ruiz, Martina Fischetti, João Paiva Fonseca
Just how different is it to work within an industry or a commercial consultancy, if you have been used to working from within academia? And how much easier is it to forge collaborations if you have experience of both sides? A panel of people who have made the move will share their experiences and lessons learned. The differences may not be the ones you expect.
TD-46 Success, failure and the factors that influence outcomes
Panellists: Dennis Huisman (Erasmus University Rotterdam & Netherlands Railways), Frédéric Gardi (Hexaly), Alberto Delgado (Maersk)
The speakers on this panel all have wide experience of working with stakeholders to deliver projects putting OR into practice. In the first part of the session, each speaker will talk about their experiences and learnings, and how these have shaped their own views on which factors influence whether an application could be (or will be?) a success, a failure or somewhere in-between. We will then invite audience questions, and open up the discussion: do these sound familiar? What are your own experiences? What lessons would you like to share on how to address them?
WA-46 Not the Excellence in Practice Award: lifting the lid on practice
Panellists include (TBC): Alberto Franco, Karsten Kieckhafer, Gilberto Montibeller
Submissions for the Excellence in Practice award naturally focus on the criteria being judged: the originality of the methodology, the scientific excellence, the impact, and so on. These usually tell an impressive story.
But anybody who has been engaged in a practical application will know that this only gives part of the picture - to get to the stage of an implemented application, there will have been many hurdles along the way: getting all the data, making sure it is correct, convincing all the stakeholders, overcoming fears and doubts of the people who have to change their way of working, and so forth.
In this session, past finalists from the Excellence in Practice award will give a short review of these aspects of their projects, considering how they dealt with the challenges and identifying key lessons. This will be followed by questions and discussion with the audience on the key issues and how we address them.
WB-46 Practical Skills: Knowledge Sharing Tutorial -2-
Expand your skills with these two 45' practical sessions: Make Easy Changes, Make Changes Easy: A Tutorial On OR Coding Quality; followed by Tricks From The Trade For Large Scale Optimisation.
(i) Make Easy Changes, Make Changes Easy: A Tutorial On OR Coding Quality (Michele Quattrone, Air Liquide)
The implementation of well-thought-through quality strategies on the software lifecycle is becoming a requirement for many software developments, even for the early mock-up/prototyping phases. In this 45-minute tutorial, backed with a practical example, we will discuss together how our coding can benefit from this strong push, and what we can cherry-pick from hyping buzzwords such as DevOps, Continuous Integration, Continuous Deployment, MLOps...in practice!
(ii) Tricks from the Trade for Large-Scale Optimization in a practical context (Andreas Schmitt, Zalando SE)
Have you had to face up to optimization challenges, where you have had to develop and deploy heuristics in order to implement models into production environments? This session will use the speaker's experience of a successful application as a starting point for discussion.
Practical optimization problems often quickly increase in scale and complexity, necessitating tailored solution approaches. Our talk describes heuristics designed to work along-side a Lagrangian decomposition method, leading to almost-optimal solutions. These have been developed in the context of the frequently applied ’predict-then-optimize’ paradigm, more specifically for markdown pricing strategies in the online fashion industry. We present empirical evidence for the heuristics’ effectiveness, drawing on pricing applications at Zalando SE.
This exposition will serve as a starting point for discussions: Participants will be encouraged to explore questions such as: What are your experiences with similar optimization challenges? How do you approach heuristic development in their work? What are effective approaches for deploying such models into production environments and measuring their impact?
WC-46 Which solver?
Participants (TBC): Johan Kellerth Fredlund (Jeppesen), Dr Faheem (Amazon), Andreas Westerlund, Tobias Achterberg (Gurobi), Timo Berthold (FICO), Fred Gardi (Hexaly)
Practitioners who use MIP have a great choice of possible softwares. The huge choice can make it harder to make decisions; and that’s where this session should be invaluable.
Bringing together representatives from leading stand-alone MIP-solvers: Gurobi, Xpress and Hexaly with representatives of heavy users of one or more of these we try to answer questions like: Why use a 3rd party solver instead of specialised heuristics? What problems have companies experienced in this approach? And of course… which solver is the best in practice?
This is a reprise of a session first held in 2018. Since then, the landscape and the power of solvers has changed immeasurably, and we will be reviewing just how users can make the most of this.
WD-46 Human Centric Analytics (HCA): How To Foster The Human Centred Development Of Analytics That Augment Human Work Effectively.
(Christina Phillips, LBS, Liverpool John Moores University)
In situations of high uncertainty and high human involvement, it can be difficult to implement analytical tools; on the other hand, those tools can serve to reduce and define the uncertainty. To achieve this level of use the human actors involved need to be empowered to understand and utilise the data that they produce and use. This requires both learning and experiment in a safe and creative space as mathematical tools, data and humans come together.
In this workshop we will use examples from industry interventions to introduce the ideas and concepts around HCA. We will discuss what works and what doesn’t, and why that is, and develop skills around choosing the right structuring methods and analytical approaches for differing situations.
Design is a key part of the process, and we will look at ways to keep in mind the iterative and empathetic journey that this helps to facilitate, while also looking at ways to integrate HCA into common working practices such as lean projects.
LIGHTNING TALKS
(The list below is not necessarily in presentation order. Details are subject to last-minute changes)
Laura Albert (University of Wisconsin-Madison): Optimizing public safety interventions to fight the opioid crisis
Worldwide, more than 500,000 deaths per year are attributable to opioid usage. The opioid crisis has strained hospitals, treatment facilities, and social service agencies due to the enormous resources and procedures needed to respond to the crisis. Public safety agencies often act as first responders, and as a result, they can play a critical role in mitigating the opioid crisis if they leverage OR and analytics. This talk overviews these opportunities, including arrest and hospital diversion, crisis response paradigms, drug interdiction, and harm reduction strategies.
Laura is currently INFORMS Past President. She is a professor of Industrial and Systems Engineering specializing in discrete optimization problems arising from applications in homeland security, disaster management, emergency response, public services, and healthcare. Her Punk Rock OR blog has brought joy to many of us in OR.
Paul Bouman (Erasmus University), Insights from a shared academic office space experiment
In the wake of the pandemic, increased working from home caused Erasmus University Rotterdam to push towards flex-working for all staff. While creating many flex-working offices theoretically maximizes sharing, this is typically disliked by employees. We tried a different approach: using an optimization model to divide a department into clusters (groups of 7 to 9 employees that can coordinate their office use as they prefer). Employees became active agents in the subsequent office arrangements. This talk discusses behavioural insights derived from desk reservation data of a 90-employee academic department working with this concept.
Paul is an associate professor at the Erasmus University Rotterdam, the Netherlands. He enjoys working on algorithms and optimization. He is interested in the impact of these techniques on the local scale of users as well as the collective system.
Héctor Cancela (Universidad de la República (Uruguay)), Delivering happiness for Uruguayan housing cooperative members
In Uruguay, housing cooperatives deliver habitational solutions to a significant part of the population. After real-state is built, individual units are assigned to cooperative members. This assignment has been traditionally performed by a random draw, without taking into account user preferences (such as sunlight orientation, height, etc.). In this talk, we share the results of a project involving students, graduates, users and faculty, which resulted in a mathematical programming based tool for performing the assignment while taking into account these preferences and thereby improving the happiness of the cooperative members.
Héctor holds a PhD. degree in Computer Science from the University of Rennes 1, INRIA Rennes, France, and a Computer Systems Engineer degree from the Universidad de la República, Uruguay. He is currently Full Professor at the Operations Research Department, Instituto de Computación, Facultad de Ingeniería, Universidad de la República (Uruguay).
Gökhan Ceyhan (JustEat Takeaway), Delivering meals, fast: learning optimal courier assignment at JET
On-demand delivery platforms manage a courier fleet aiming to deliver the orders as quickly as possible and in a cost-efficient way. A core component of the platform is the courier assignment algorithm assigning incoming orders to couriers in real-time. This talk describes a project to learn the optimal assignments via machine learning, using the characteristics of the problem instance and the feasible decision space. It also looks at how well we can expect this method to work in practice, in terms of solution quality and run times.
Gökhan has been working as an OR Scientist at Just Eat Takeaway.com since 2021, focused on tactical and operational level logistics problems. Previously, he worked on different optimization problems in the energy and aviation sectors. His work covers mixed-integer optimization, multi-objective optimization, and recently, machine learning.
Ahmet Çınar (Trendyol), Navigating the Challenges of Operations Research in Sequential Decision-Making for E-commerce
When customers place orders through e-commerce applications, it initiates a sequence of decision-making processes. These range from selecting the warehouse that will fulfill the order to choosing the mode of transportation for delivery. In this talk, we will explore the various decision-making points and challenges in the realm of e-commerce.
Ahmet has a diverse educational background with a BSc in Industrial Engineering, PhD in Operations Management, and BA in Law. With 15 years of experience in OR and Data Science, he currently works at Trendyol. In addition to professional pursuits, he enjoys playing traditional instruments like ney and kanun.
Claudio Gambella (FICO), Assigning customers to financial treatments: an Action Effect approach to predict customer behavior
The FICO Decision Optimizer (DO) application is an optimization software package to perform optimal assignment of treatments to a portfolio of customers. The DO models depend on structural inputs (e.g., portfolio or customer attributes) and uncertain inputs (predictable target values, such as the impact of treatments on customers). Developing Predictive Causal Models usually requires additional tools and knowledge for a Business Analyst. To simplify the user experience, we developed an Action Effect (AE) approach to predict the impact of the treatment assignment on every customer of the portfolio.
Claudio has been an Operations Research Modeler at FICO since 2021. He is interested in modelling and solving optimization problems and machine learning tasks arising from real-world applications. Before joining FICO, he worked with IBM Research Ireland. He has received an AIRO Best Paper and IBM Research Awards.
Caroline Jagtenberg (Vrije Universiteit Amsterdam), Heroes Next Door: Empowering Community First Responders
Traditional emergency services can be helped by so-called Community First Responders (CFRs): nearby, trained volunteers dispatched via a smartphone app. The volunteer covers those first crucial moments and makes a tremendous difference by e.g. resuscitating a cardiac arrest patient or preventing the spread of a fire, before the emergency services arrive. Several such systems exist worldwide, but the OR community has only recently begun to model them. Crucial is the stochasticity in the volunteer behavior: their proximity to incidents and willingness to accept determines their effectiveness. We discuss ongoing work that includes a collaboration with 5 practitioners in 3 countries.
Caroline is an assistant professor at Vrije Universiteit Amsterdam. She also worked part time in various industry jobs. Her research focuses on emergency services: in ambulances, fire fighters and police, she finds collaborators and novel research questions. Her data-driven models help to improve their efficiency, fairness and cost effectiveness.
Johan Kellerth Fredlund (Jeppesen), Flight retiming in the Aircraft Routing Optimiser
Aircraft Routing is an optimizer developed by Jeppesen used by many airlines around the world to assign flights to aircraft and optimize routes. This is done while satisfying critical constraints like minimum turn times and maintenance events. For certain scenarios the need for retiming is needed to ensure that the solution is feasible and of high quality. The case for introducing flight retiming support to the optimizer will be discussed during this presentation, as well as some of the use cases and how it benefits the airlines.
Johan got his MSc in Control Theory at Lund University in 2017 and is working as a Data Scientist at Jeppesen, a Boeing Company. He and his team develop fleet optimizers like Aircraft Routing, Tail Assignment and Flight Scheduling.
Mirjam Meijer (YW4OR, Kühne Logistics University), Challenges and opportunities in high-tech manufacturing
In high-tech manufacturing, timeliness is essential for smooth operations. Whether it's lithography machines or aircraft, the assembly of complex systems depends on the availability of every component at the right moment. Given the large number of components required for building such high-tech systems, this is a difficult problem, which becomes even more challenging when considering specific characteristics of high-tech supply chains. On the other hand, the structure of these supply chains gives rise to some opportunities. We will consider these challenges, opportunities, and their consequences for practice and research.
Mirjam is an assistant professor of operations management at the Kühne Logistics University (KLU) in Hamburg, Germany. Before joining KLU, she obtained her PhD from Eindhoven University of Technology, the Netherlands. In 2022 she was selected as one of the YoungWomen4OR by the EURO WISDOM Forum.
Richard Oberdieck (Banking Circle), What OR can learn from putting Large Language Models (LLMs) into production
Developing a new optimization model is arguably the most exciting part of being an OR professional. However, once a passable version of a model is created, how can it be put to use in a production environment? This is not taught at universities, nor is it a focus of the OR professionals. This talk will argue that it should be. The author will use a recent example where he and his colleagues deployed a Large Language Model into a production environment of a critical part of a banking system to show what OR can learn from this new technology.
Richard Oberdieck obtained his PhD from Imperial College London. He then worked for the offshore wind company Ørsted, before moving to the technical sales team at Gurobi Optimization. Now, he is a Lead Data Scientist at Banking Circle, a fintech bank, and is based in Copenhagen, Denmark.
Hans Schwartzlose, Kong You Liow (Ørsted) Scheduling in Offshore Wind: Challenges and Opportunities
In the dynamic environment of offshore wind power operations, Ørsted faces the challenge of limited resources and unpredictable weather impacting access to wind farms. Effective scheduling tools are essential to manage these uncertainties, ensuring maintenance and operations are conducted efficiently and within budget. This presentation underscores the importance of robust scheduling plans that accommodate the variable nature of offshore activities, highlighting the need for flexibility in Ørsted's operational strategy.
Kong You is a decision scientist at Ørsted with a BSc in Theoretical Physics from Imperial College London and a PhD in Astrophysics from University of Exeter. Hans, a senior decision scientist at Ørsted, holds a Master's in Mathematical Modelling from DTU and has extensive data science expertise.
Carolina van Weerd (TNO). Art vs science
While no one is able to predict the future, strategic foresight is critical in today’s complex and changing world. Strategic foresight is a discipline in which expert judgement remains central, but TNO’s VISTA (VIew on STrategic Anticipation) adds more analytical rigour (‘science’) to the much needed creativity (‘art’) in strategic foresight. The emphasis should be on a systems perspective: taking an integral approach, by not only looking at trends in isolation, but to explicitly relate trends form different domains with one another. In the words of Dwight D. Eisenhower: “plans are worthless, but planning is everything.”
Carolina is a researcher and consultant at TNO Defence, Safety and Security in the Netherlands. She has a background in International Relations and Military Strategic Studies and specializes in strategic foresight, with a focus on the potential impact of technology on security and defence.
