08/2026 BIFOLD Colloquium
June
23,
2026 
16:00
- 17:00
Scaling SQL's Expressiveness with High-Performance User-Defined Functions
Abstract:
User-defined functions (UDFs) extend SQL with custom computation, enabling seamless integration of declarative querying and functional capabilities for application-specific logic, including complex analytics, data science, and machine learning workloads. While this additional expressiveness is essential for modern data applications, it introduces significant challenges for query optimization and execution, as user-defined code often falls outside the traditional optimization framework of relational systems. In this talk, we will examine how research and commercial systems bridge this gap through techniques ranging from algebraic, cost-based optimization to low-level, physical query optimization, compilation, and execution, including methods such as vectorization, parallelization, tracing JIT compilation, and operator fusion. We will discuss support for common UDF types, such as scalar, aggregate, and table UDFs, and show how these techniques enable high-performance execution without sacrificing programmability. Finally, we will highlight emerging opportunities for scaling user-defined computation using heterogeneous accelerators, including GPUs, and for integrating rich user-defined logic into streaming and AI-centric data processing systems.
Short-bio:
Alkis Simitsis is a Research Director at Athena Research Center. In the past, he held positions with HP/HPE Labs, Micro Focus, Unravel Data, and IBM Research, including Chief Scientist, Systems Architect, and Principal Research Scientist. Alkis brings over 20 years of experience across both startup and corporate environments, developing innovative information and data management solutions as well as enterprise-grade products. His work spans scalable big data infrastructure, data-intensive analytics, information management, business intelligence, massively parallel processing, distributed and column-store databases, security analytics, and cloud computing. Alkis holds 45 U.S. and 1 European patents, has published 130+ papers in refereed international journals and conferences (top publications cited 8000+ times, h-index: 47), and frequently serves in the organization and program committees of top-tier international scientific conferences. His most recent service includes General co-chair for VLDB 2027, PC co-chair for IEEE ICDE 2026 and EDBT 2025, associate editor for ACM SIGMOD 2027/2026, PVLDB 2026/2025/2023, IEEE ICDE 2027/2023, and IEEE TKDE, and he is on the editorial board of ACM/IMS J. of Data Science, VLDB Journal, and Elsevier DKE. Alkis is a recipient of the ACM DOLAP 2023 Test-of-Time Award, several best paper/demo awards (VLDB 2025, IEEE ICDE 2024, EDBT 2024, ACM CIKM 2020, ACM SIGMOD 2014 and 2012) and service awards (PVLDB 2025/2023, IEEE ICDE 2023, EDBT 2024/2023, ACM SIGMOD 2021).