[ECSS 2025] Presentation of the Keynote Session (28 October 2025)

European Informatics Leaders Summit ECSS 2025 The European Voice of Informatics Research and Education Keynote Session Rennes, 28 October, 2025 JOHANNES KEPLER UNIVERSITY LINZ Altenberger Straße 69 4040 Linz, Austria jku.at How to prepare for a new computing paradigm? Quantum Software ante portas Manuel Wimmer Institute of Business Informatics - Software Engineering (BISE) [email protected] What can we really expect from QC? (unfortunately) Nothing New ! Every QC-solvable problem can be solved on a classical machine (with enough time). Actually, we currently simulate QC. But… … for some problems, certain quantum algorithms have exponentially better runtime complexity. These algorithms are hard to find: “quantum computers are very difficult to reason about using classical intuition” Shor, Peter W. "Why haven't more quantum algorithms been found?" Journal of the ACM (JACM) 50.1 (2003): 87-90. 3 The Potential of QC Source: Hidary (2019). Quantum Computing: An Applied Approach, Springer 2021 ●Many complex problems are intractable for classical computing ○Exponentially growing search spaces ●Classical computers are reaching physical limitations ●Which problems can quantum computers solve faster than classical computers can? ●Aim: QC algorithms which provide polynomial/exponential speedup 4 Where QC is expected to excel? Quantum Simulation: Natural modeling of quantum systems (chemistry, materials, molecular reactions) by following Feynman’s original motivation. Optimization Problems: Variational quantum algorithms (VQE, QAOA) and annealers tackle combinatorial or NP-hard problems. Search and Sampling: Quadratic speed-up in unstructured search (Grover’s algorithm); quantum Monte Carlo for complex distributions. Cryptanalysis: Factoring and discrete logarithms (Shor’s algorithm) → impact on RSA. Machine Learning & Pattern Discovery: Quantum kernel methods, amplitude encoding, and quantum feature spaces for high-dimensional data. Hybrid Workflows: Acceleration of subroutines (e.g., optimization or linear algebra) inside classical pipelines. 5 Example: Quantum Combinatorial Optimization Felix Gemeinhardt, Antonio Garmendia, Manuel Wimmer, Benjamin Weder, Frank Leymann: Quantum Combinatorial Optimization in the NISQ Era: A Systematic Mapping Study. ACM Comput. Surv. 56(3), 2024 •QCO aims to find the set of discrete decision variables that minimize a certain constrained or unconstrained optimization function. •Problems are usually non-convex (causing many local optima) and are often characterized by an exponentially growing search space 6 Example: Quantum Combinatorial Optimization Felix Gemeinhardt, Antonio Garmendia, Manuel Wimmer, Benjamin Weder, Frank Leymann: Quantum Combinatorial Optimization in the NISQ Era: A Systematic Mapping Study. ACM Comput. Surv. 56(3), 2024 •QCO aims to find the set of discrete decision variables that minimize a certain constrained or unconstrained optimization function. •Problems are usually non-convex (causing many local optima) and are often characterized by an exponentially growing search space 7 Example: Quantum Combinatorial Optimization Felix Gemeinhardt, Antonio Garmendia, Manuel Wimmer, Benjamin Weder, Frank Leymann: Quantum Combinatorial Optimization in the NISQ Era: A Systematic Mapping Study. ACM Comput. Surv. 56(3), 2024 •QCO aims to find the set of discrete decision variables that minimize a certain constrained or unconstrained optimization function. •Problems are usually non-convex (causing many local optima) and are often characterized by an exponentially growing search space 8 What to expect from the rest of this talk? •Quick outline of QC ◦How does the technology work? ◦What can it do? •Quantum Software in addition to Classical Software ◦Where are we? ◦Current challenges •How to approach Quantum Software research? ◦Some insights how we approach this area •What’s next? 15 Quantum Technologies: What are we speaking about? Quantum Communication Quantum Sensing Quantum Computing Quantum Software is required for all these technologies, but especially for Quantum Computing 16 What is Quantum Computing? Quantum computers process information using the laws of quantum physics, not classical Boolean logic. While a classical computer is in one of 2^N states at a time (where N is number of bits), a quantum computer is in probability distribution of 2^N states. Programming a QC is not about transforming one state through operations a computer can perform faster than a human. It is about manipulating a distribution over all possible solutions to a problem in a meaningful way. 17 Basic Concepts - Superposition Classic bits: two states (0, 1) Quantum bits (Qubits): represent a probability distribution between 0 and 1 Upon measurement, states collapse to 0 or 1 ψ=α ∗ |0〉 + 𝛽 ∗ |1〉,where |α|2+ |β|2= 1 Bloch Sphere Similar to classical bit 0,1 →ۧ |0 , ۧ |1 Can also be a mixture →superposition 18 Basic Concepts - Entanglement Correlation between probability distributions of qubits ◦Probabilities of two qubits can be entangled E.g.,: Bell States (completely entangled): ◦ۧ |Ψ+=𝟏 𝟐ۧ |𝟎𝟎 + 𝟏 𝟐ۧ |𝟏𝟏 ◦When one qubit is measured, the shared superposition collapses. Right after measurement, both qubits have the same value Entanglement is an important building block for achieving speed-ups in quantum algorithms Side-note: Linking two quantum computers through entanglement gives exponential speedup Jozsa, R., & Linden, N. (2003). On the Role of Entanglement in Quantum-Computational Speed-Up. Mathematical, Physical and Engineering Sciences, 459(2036). 19 50% probability of measuring |00> 50% probability of measuring |11> 0% probability of measuring |01> or |10> Basic Concepts - Quantum Parallelism Because our system is in a multitude of states at once (superposition), a single operation can affect all of these states at once. There is no free lunch! Due to the laws of quantum physics, quantum operations have to be reversible. Most operations from classical computing are not reversible (e.g., AND, OR, …), and therefore, cannot directly be executed on a QC. 20 Basic Concepts - Measurement Measurement destroys superposition •Non-reversible quantum operation Intermediate states of the quantum system are not accessible Copying is not possible, see no-cloning theorem •To reproduce states, repeated state computation and measurement required 21 Wootters, W., Zurek, W. A single quantum cannot be cloned. Nature 299, 802–803 (1982) What does a quantum program look like? Recall: Classical Logic Circuit Diagram Input Output Logic Gates (And/Or/Xor/...) Bits Bits Computation 22 What does a quantum program look like? Quantum Gate actions on the qubits •X (Pauli-X): bit flip •Z (Pauli-Z): phase flip •H (Hadamard): creates superposition •… A unitary matrix Uacting on a qubit 𝜓′= 𝑈|𝜓ۧ Gates are reversible (important property) Typical Problem: “Create a circuit that does …” Current Status: Quantum Circuit Input/ Output qubits Quantum Operations Gate Set {X, Z, H, …} Initialization Computation Measurements 23 What does a quantum program look like? Quantum Gate actions on the qubits •X (Pauli-X): bit flip •Z (Pauli-Z): phase flip •H (Hadamard): creates superposition •… A unitary matrix Uacting on a qubit 𝜓′= 𝑈|𝜓ۧ Gates are reversible (important property) Typical Problem: “Create a circuit that does …” Current Status: Quantum Circuit Input/ Output qubits Quantum Operations Gate Set {X, Z, H, …} Initialization Computation Measurements 24 Abstraction levels Models, No-Code, 4GL, 5GL, … Python + Frameworks Java, …: objects C++: objects C: arrays, data types, … Assembler Language Machine language (00100101…) ?Models, No-Code, 4GL, 5GL? IBM Quantum Composer (qubits) Python + Frameworks (qubits) … Circuits before transpilation (qubits) Circuits after transpilation (qubits) Pulse-level, … Classical Software Quantum Software 31 O. Di Matteo et al. "An Abstraction Hierarchy Toward Productive Quantum Programming," IEEE International Conference on Quantum Computing and Engineering (QCE), 2024. Meta-meta-model Meta-model Model Hybrid Application UML Models <<conformsTo>> <<conformsTo>> Classical Computer Classic Computing Components Classical/Quantum Application Quantum Computer Quantum Computing Components UML MOF And what about Modeling Languages? 32 ? ! Q-UML 2020 QuanUML 2025 … Metamodeling Stack DSMLs Models UML Models UML DSMLs MOF And what about Modeling Languages? 33 ? Felix Gemeinhardt, Antonio Garmendia, Manuel Wimmer: Towards Model-Driven Quantum Software Engineering. Q-SE@ICSE, 2021. <<conformsTo>> <<conformsTo>> Metamodeling Stack Hybrid Application Classical Computer Classic Computing Components Classical/Quantum Application Quantum Computer Quantum Computing Components Meta-meta-model Meta-model Model DSMLs Models UML Models Quantum Models UML QuantumML DSMLs MOF And what about Modeling Languages? 34 ! ? Felix Gemeinhardt, Antonio Garmendia, Manuel Wimmer, Robert Wille. A Model-Driven Framework for Composition-Based Quantum Circuit Design. ACM Transactions on Quantum Computing 5(4), 2024. Meta-meta-model Meta-model Model Hybrid Application Classical Computer Classic Computing Components Classical/Quantum Application Quantum Computer Quantum Computing Components Metamodeling Stack <<conformsTo>> <<conformsTo>> Reality Check: Multitude of Challenges/Opportunities QC community focus •More and better qubits (hardware) •Demonstrating quantum advantage (algorithms) However, software is the next bottleneck •Abstraction gap: gate-level (like assembly!) •No best practices: Testing, debugging, verification •Tooling immaturity: Limited IDE support, e.g., profilers, debuggers, validators, … •Skills gap: Quantum physicists ≠ software engineers Opportunity for computer scientists (Non-exhaustive List!) ✅Model-driven engineering ✅Search-based software engineering ✅Testing and verification ✅Software processes ✅Service Oriented Computing ✅Software Architecture ✅Quantum AI Juan Manuel Murillo et al.: Quantum Software Engineering: Roadmap and Challenges Ahead. ACM Trans. Softw. Eng. Methodol. 34(5): 154:1-154:48 (2025) 35 Integrated Quantum Computer Qubits and Controls Classical Computer Quantum Algorithm Library Transformer Compiler Q Algo High-Level Quantum Program Low-Level Quantum Program Raw Data Post-Processor (e.g., Error Mitigation) Computation Result (Solution) Computational Problem Quantum Computer QSE Focus at BISE (JKU Linz) 36 (2) SBSE Quantum Circuit Synthesis and Debugging with Search-based Software Engineering Composition-based Quantum Circuit Design with Model-Driven Software Engineering (1) MDSE Integrated Quantum Computer Illustration based on Proctor et al. Benchmarking quantum computers. Nat Rev Phys 7, 105–118 (2025). Area #1: MDSE 4 QSE Methodology 37 https://github.com/jku-win-se/composition-quantum-circuit Felix Gemeinhardt, Antonio Garmendia, Manuel Wimmer, Robert Wille. A Model-Driven Framework for Composition-Based Quantum Circuit Design. ACM Transactions on Quantum Computing 5(4), 2024. Area #1: MDSE 4 QSE Metamodel for Quantum Circuits Methodology 38 https://github.com/jku-win-se/composition-quantum-circuit Felix Gemeinhardt, Antonio Garmendia, Manuel Wimmer, Robert Wille. A Model-Driven Framework for Composition-Based Quantum Circuit Design. ACM Transactions on Quantum Computing 5(4), 2024. Area #1: MDSE 4 QSE Metamodel for Quantum Circuits Methodology 39 https://github.com/jku-win-se/composition-quantum-circuit Felix Gemeinhardt, Antonio Garmendia, Manuel Wimmer, Robert Wille. A Model-Driven Framework for Composition-Based Quantum Circuit Design. ACM Transactions on Quantum Computing 5(4), 2024. Metric-based Evaluation (Comparison with IBM Quantum Composer) Area #2: Quantum Circuit Synthesis (QCS) 40 Goal: Find a quantum circuit that implements a desired behavior. Circuit RQ1: better diversity RQ2: Hybrid has higher overlap + faster convergence Hybrid Non-Hybrid 47 Challenge: Hard to navigate search space Hybrid Quantum State Synthesis 48 RQ3: higher overlap, but more costly circuits Hybrid Non-Hybrid Challenge: Hard to navigate search space Hybrid Quantum State Synthesis Core Idea •Use existing information as “seed” for searching for better versions Results •Can fix buggy and optimize inefficient quantum circuits Problem •Sometimes we already have a scaffold or almost correct quantum circuit 49 Challenge: Hard to navigate search space Hybrid Quantum Debugging and Optimization Felix Gemeinhardt, Stefan Klikovits, Manuel Wimmer: GeQuPI: Quantum Program Improvement with Multi-Objective Genetic Programming. J. Syst. Softw. 219 (2025) Debugging capabilities? •Hybrid can optimize ◦for specific input states ◦with Non-Hybrid & Fixed sometimes only equally good solutions are found •Hybrid can repair arbitrary inputstate programs ◦Non-Hybrid & Fixed have problems Debugging capabilities: number of runs per category and use case. (Hybrid /Non-Hybrid /Fixed). 50 Optimization capabilities? •All approaches can optimize •Hybrid performs (almost) consistently better •Hybrid improves by 35% •Hybrid compared to “standard approach” (Qiskit built-in optimizer): ◦optimize in significantly more cases ◦and higher on average Optimization capabilities (Hybrid /Non-Hybrid /Fixed). 51 Lesson Learned: Simulation of the candidate circuits consumes majority of the execution time of the discussed approaches! Ongoing work: Simulation with Caching 52 Problem -QCS (often) operates on circuit populations (e.g., when using GAs) -Populations (often and on purpose) contain redundancy Core Idea -Identify recurring gate patterns and aim for reuse of the results →reduce computations Results −Clear reduction in simulation time (depending on redundancy and qubits) > 50% time savings in GAs Ongoing work: Divide & Conquer Simulation 53 Core Idea −Use Reinforcement Learning to split synthesis problem and solve subproblems independently Problem −Cost of simulation grows exponentially in number of qubits Results −Agent was able to divide in > 75% of all cases −Much fewer gates needed for separation than for synthesis of the circuits Summary –Lessons Learned 54 Adaptation of Methodologies •Possible, but requires completely novel treatments for particular subproblems •Requires interdisciplinary exchange and new skills for researchers Technological Spaces • Integrating QC into “your” technologies (we did this for the EMF ecosystem) •Allows to reuse existing tooling (which is often defined on the meta-level) • Model transformations, code generation, validation, … Quantum Europe Strategy „However, this is still insufficient to meet the projected demand from EU’s startups and industry, which faces major shortages of professionals with relevant applied skills.“ „Shortages are most critical in applied fields, including quantum software engineering, system integration, and quantum cybersecurity, slowing the commercialisation path for EU-based startups and scaleups.“ Taken from: https://digital-strategy.ec.europa.eu/en/policies/quantum https://digital-strategy.ec.europa.eu/en/library/quantum-europe-strategy Area 5 - Quantum Skills: Building a diverse, world-class workforce through coordinated education, training, and talent mobility across the EU 55 Academic Study Programmes Characteristics •Heterogeneous landscape of programmes •Focus on MSc degree (typically building on CS or physics) •Industry Partners: IBM, Microsoft, Google, AWS, … •Include Q-Programming, more advanced software engineering topics rarely embedded within broader QC curricula Notable Examples •Quantum Master Barcelona (incl. Quantum Software track) •Aalto: BSc →MSc →PhD in Quantum Technology •Gdansk: International MSc in QIT What is Missing •Dedicated Programme for Quantum Software (Software Specialisation only at KAIST, SDU, UTS) •Shared understanding on “how to design such a programme” 20+ 50+ 15+ 13+ 10+ 8+ 100+ Programmes Worldwide 56