Applied Research at Inter&co

The Quantum Approximate Optimization Algorithm (QAOA) is a leading candidate for solving Quadratic Unconstrained Binary Optimization (QUBO) problems. However, its performance on Noisy Intermediate-Scale Quantum (NISQ) devices is hindered by limited qubit connectivity, which increases circuit depth. We introduce a hardware-aware QAOA framework, using a Mixed Integer Semidefinite Programming (MISDP) formulation to find an approximate cost Hamiltonian compatible with hardware topology. To solve the MISDP, we propose heuristics based on spectral properties of the hardware graph. We apply this framework to the index tracking problem, and our numerical results demonstrate that this approach yields high-quality solutions.

In the graph-theoretic approach to contextuality, a key problem is to give a foundational justification for quantum theory from some simple physical principles. This was already solved for perfect graphs. We investigate whether assuming this to hold for minimally imperfect graphs implies it for all imperfect graphs. We confirm this hypothesis for a large family of graphs, G, which contains all minimally imperfect graphs and is closed under complement, disjoint union, and addition of twins. Our proof relies on a new characterization of the theta body of a graph under the addition of twins.

We present the Generative Quantum Approximate Optimization Algorithm (GQAOA), a method for applying classical generative models to optimize QAOA circuit parameters. Our work is inspired by the Generative Quantum Eigensolver (GQE), which trains a classical generative model to produce a sequence of quantum gates to generate quantum states that minimize the energy of a Hamiltonian. Our method adapts this framework to optimize the QAOA circuit parameters, given an Ising Hamiltonian as input that represents a real-world portfolio optimization problem.

We introduce CodeEvolve, an open-source framework that combines large language models (LLMs) with evolutionary search to synthesize high-performing algorithmic solutions. CodeEvolve couples an islands-based genetic algorithm with modular LLM orchestration, using execution feedback and task-specific metrics to guide selection and variation. Exploration and exploitation are balanced through context-aware recombination, adaptive meta-prompting, and targeted refinement of promising solutions. We evaluate CodeEvolve on benchmarks used to assess Google DeepMind’s AlphaEvolve, and include direct comparisons with popular open-source frameworks for algorithmic discovery and heuristic design. Our results show that CodeEvolve achieves state-of-the-art (SOTA) performance on several tasks, with open-weight models often matching or exceeding closed-source baselines at a fraction of the compute cost.

This article presents an innovative project for a Central Bank Digital Currency (CBDC) infrastructure. Focusing on security and reliability, the proposed architecture employs post-quantum cryptography (PQC) algorithms for long-term security, can be integrated with a Trusted Execution Environment (TEE) to safeguard confidentiality, and uses Distributed Ledger Technology (DLT) to promote transparency and tamper resistance. We experimentally evaluate CRYSTALS-Dilithium, Falcon, and SPHINCS+ integrated into Hyperledger Besu, both inside and outside an Intel SGX TEE. CRYSTALS-Dilithium-2 combined with classical secp256k1 signatures achieves the shortest execution times, reaching 1.68 ms for block signing without TEE and 0.5 ms for signature verification.

Money laundering has become one of the most relevant criminal activities in modern societies, as it causes massive financial losses for governments, banks and other institutions. Detecting such activities is among the top priorities when it comes to financial analysis, but current approaches are often costly and labor intensive partly due to the sheer amount of data to be analyzed. Hence, there is a growing need for automatic anti-money laundering systems to assist experts. In this work, we propose DELATOR, a novel framework for detecting money laundering activities based on graph neural networks that learn from large-scale temporal graphs. DELATOR provides an effective and efficient method for learning from heavily imbalanced graph data, by adapting concepts from the GraphSMOTE framework and incorporating elements of multi-task learning to obtain rich node embeddings for node classification. DELATOR outperforms all considered baselines, including an off-the-shelf solution from Amazon AWS by 23% with respect to AUC-ROC. We also conducted real experiments that led to the discovery of 7 new suspicious cases among the 50 analyzed ones, which have been reported to the authorities.