Research

Foundations of next-generation data management systems

Schema-Based Query Optimisation for Graph Databases

Recursive graph queries are increasingly popular for extracting information from interconnected data found in various domains such as social networks, life sciences, and business analytics. Graph data often come with schema information that describe how nodes and edges are organized. We propose a type inference mechanism that enriches recursive graph queries with relevant structural information contained in a graph schema. We show that this schema information can be useful in order to improve the performance when evaluating recursive graph queries. Furthermore, we prove that the proposed method is sound and complete, ensuring that the semantics of the query is preserved during the schema-enrichment process. Experimental results with a complete implementation of the approach show significant performance gains for query evaluations over property graphs, with several evaluation backend. These results were presented at SIGMOD 2025 in Berlin 7.

Distributed Evaluation of Graph Queries Using Recursive Relational Algebra

We present a system, Dist-μ-RA ,for the distributed evaluation of recursive graph queries. Dist-μ-RA builds on the recursive relational algebra 2, 1 and extends it with evaluation plans suited for the distributed setting. The goal is to offer expressivity for high-level queries while providing efficiency at scale and reducing communication costs. Specifically, we propose a new approach for the evaluation of recursive algebraic terms in a distributed manner. The method enables generating independent parallel loops on the worker nodes in a cluster of machines instead of executing a global loop on the driver node. The advantage of the parallel local loops is a minimization of the amount of data shuffled between worker nodes. This reduces communication costs and significantly improves overall query evaluation time. We applied this approach to recursive graph queries on real and synthetic datasets. Experimental results on both real and synthetic graphs show the effectiveness of the proposed approach compared to existing systems. These results were presented at ICDE 2025 in Hong Kong 5 [6.1.1].

Efficient Iterative Programs with Distributed Data Collections

Big data programming frameworks have become increasingly important for the development of applications for which performance and scalability are critical. In those complex frameworks, optimizing code by hand is hard and time-consuming, making automated optimization particularly necessary. In order to automate optimization, a prerequisite is to find suitable abstractions to represent programs; for instance, algebras based on monads or monoids to represent distributed data collections. Currently, however, such algebras do not represent recursive programs in a way which allows for analyzing or rewriting them. In this paper, we extend a monoid algebra with a fixpoint operator for representing recursion as a first class citizen and show how it enables new optimizations. Experiments with the Spark platform illustrate performance gains brought by these systematic optimizations. These results have been published in the Journal of Logical and Algebraic Methods in Programming 3.

An Enterprise Marketplace for Unified Access to Multi-Cloud and Enterprise Products in a Large Banking Infrastructure

We present the design and evaluation of an Enterprise Marketplace that unifies web-based access to multi-cloud and enterprise products within a large banking infrastructure. Building such a platform poses significant technical and organizational challenges, including the generalization of diverse APIs and the accommodation of heterogeneous user profiles. Our solution enables autonomous product publishing via a no-code interface, enforces multi-layered governance to ensure security and compliance, and integrates disparate providers through a standardized API. We demonstrate how this approach enhances product quality, producer autonomy, and user experience, supported by adoption metrics and operational data from a real-world deployment. Finally, we reflect on key lessons learned and persistent challenges after a decade of production use, serving over 200,000 users. These results will be presented at ICSE 2026 in Rio de Janeiro 4.

COSMetyc: OpenStreetMap in OCaml

We introduce COSMetyc, an OCaml library for manipulating OpenStreetMap (OSM) data, a large-scale collaborative geographic database. Given the heterogeneity of usages, formats, and representations in the OSM ecosystem, COSMetyc adopts a modular design that leverages OCaml’s type system to statically guarantee data validity. The library supports importing and exporting data from multiple formats (notably GeoJSON and OSM XML) through several typed representations, provides conversions between coordinate systems, and enables efficient spatial queries. We report on the design choices underlying the library and show how OCaml features such as functors, phantom types, and polymorphic variants can be used to manage this complexity in a principled and scalable way 8.

Neurosymbolic AI

A Comparative Analysis of Neuro-symbolic Methods for Link Prediction

Link prediction on knowledge graphs is relevant to various applications, such as recommendation systems, question answering, and entity search. This task has been approached from different perspectives: symbolic methods leverage rule-based reasoning but struggle with scalability and noise, while knowledge graph embeddings (KGE) represent entities and relations in a continuous space, enabling scalability but often neglecting logical constraints from ontologies. Recently, neurosymbolic approaches have emerged to bridge this gap by integrating embedding-based learning with symbolic reasoning. This paper provides a structured review of state-of-the-art neurosymbolic methods for link prediction. Beyond a qualitative analysis, a key contribution of this work is a comprehensive experimental benchmarking, where we systematically compare these methods on the same datasets using the same metrics. This unified experimental setup allows for a fair assessment of their strengths and limitations, bringing elements of answers to following key questions: How accurate are these methods? How scalable are they? How beneficial are they for different levels of provided knowledge and to which extent are they robust to incorrect knowledge? These results have been presented at the NeSy 2025 conference in Santa Cruz 6.

On Scaling Neurosymbolic Programming through Guided Logical Inference

Probabilistic neurosymbolic learning seeks to integrate neural networks with symbolic programming. Many state-of-the-art systems rely on a reduction to the Probabilistic Weighted Model Counting Problem (PWMC), which requires computing a Boolean formula called the logical provenance. However, PWMC is #P-hard, and the number of clauses in the logical provenance formula can grow exponentially, creating a major bottleneck that significantly limits the applicability of PNL solutions in practice. We propose a new approach centered around an exact algorithm DPNL, that enables bypassing the computation of the logical provenance. The DPNL approach relies on the principles of an oracle and a recursive DPLL-like decomposition in order to guide and speed up logical inference. Furthermore, we show that this approach can be adapted for approximate reasoning guarantees, called ApproxDPNL. Experiments show significant performance gains. In particular, DPNL enables scaling exact inference further, resulting in more accurate models 9.

OntoKGE: A Framework for Injecting Ontology Rules into Knowledge Graph Embedding Training.

Link prediction is a key task in knowledge graphs that involves inferring new links between entities based on existing ones. Knowledge graph embedding models address this task by representing entities and relations as points in a geometric space and using distance or similarity functions to predict missing links. However, traditional knowledge graph embedding models are trained only on assertional facts and ignore semantic information, encoded for example in the ontology. We aim to improve the quality of knowledge graph embedding models for link prediction by leveraging prior knowledge. To this end, we propose OntoKGE, an embedding-agnostic framework that integrates a reasoning module into the training process of knowledge graph embedding models. This module is based on Datalog and derives additional training facts, enriching the training set and enhancing the learned embeddings. Experiments on multiple benchmarks show that OntoKGE improves link prediction performance across multiple knowledge graph embedding models 10.