Viktor Sanca

PhD candidate at EPFL.
Interested in systems, databases, and ML. Especially how to make them efficient and fast on modern hardware.

My research is in data management systems, machine learning, and applications. My primary focus is on using approximate query processing with high-performance analytical systems on modern hardware to enable workload-agnostic, timely, interactive, and cost-efficient insights from the growing amounts of data [2,9,11,12]. Broadly, I am interested in novel hardware architectures, hardware-software co-design, database systems, ML, and algorithms for analytics [1,3,5,9] - particularly the intersection of vector-relational analytics that results in an extended relational model. This drives my interest in using machine learning with databases, particularly in mixed data format analytics and holistic optimization for semantic processing, from complex logical to physical plans that execute on heterogeneous hardware [6,7,10,12]. Finally, modern hardware requires special consideration to achieve the best performance; thus, analyzing and optimizing algorithms for modern platforms and systems remains a natural goal [5,8,9,10].

The latest publications are available at my Google Scholar profile.

Thesis

EPFL, 2024, Efficient Approximate Analytics via Adaptive Context-Conscious Query Processing. Link

Advisor: Prof. Anastasia Ailamaki

Committee: Prof. Anne-Marie Kermarrec, Prof. Carsten Binnig, Dr. Justin Levandoski

Publications

  1. Viktor Sanca and Anastasia Ailamaki. 2024. Efficient Data Access Paths for Mixed Vector-Relational Search. DAMON@SIGMOD’24. Link

  2. Viktor Sanca and Anastasia Ailamaki. 2024. Efficient and Reusable Lazy Sampling, SIGMOD Record, Link

  3. Viktor Sanca, Manos Chatzakis, Anastasia Ailamaki. 2024. Optimizing Context-Enhanced Relational Joins, ICDE, Link

  4. Viktor Sanca and Anastasia Ailamaki. 2024. Efficient Model-Relational Data Management: Challenges and Opportunities. TKDE Special Issue on Best and Innovation Papers, 2024. Link

  5. Viktor Sanca and Anastasia Ailamaki. 2023. Context-Enhanced Relational Operators with Vector Embeddings. Under submission. Preprint

  6. Viktor Sanca and Anastasia Ailamaki. 2023. Post-Moore’s Law Fusion: High-Bandwidth Memory, Accelerators, and Native Half-Precision Processing for CPU-Local Analytics. ADMS@VLDB’23. Link

  7. Viktor Sanca and Anastasia Ailamaki. 2023. E-Scan: Consuming Contextual Data with Model Plugins. CDMS@VLDB’23. Link

  8. Stefan Igescu, Viktor Sanca, Eleni Zapridou, and Anastasia Ailamaki. 2023. Improving K-means Clustering Using Speculation. AIDB@VLDB’23. Link

  9. Hamish Nicholson, Andreea Nica, Aunn Raza, Viktor Sanca, and Anastasia Ailamaki. 2023. Chaosity: Understanding Contemporary NUMA-architectures. To appear at the TPC-TC@VLDB’23. Preprint

  10. Viktor Sanca, Periklis Chrysogelos, and Anastasia Ailamaki. 2023. LAQy: Efficient and Reusable Query Approximations via Lazy Sampling. In Proc. ACM Manag. Data 1(2): 174:1-174:26 (2023). Link - Presented at SIGMOD’23 - SIGMOD Research Highlight Award

  11. Viktor Sanca and Anastasia Ailamaki. 2023. Analytical Engines With Context-Rich Processing: Towards Efficient Next-Generation Analytics. In 2023 IEEE 39th International Conference on Data Engineering (ICDE). Link - Special track vision paper, presented at ICDE’23, invited for TKDE Special issue for Best and Innovation Papers.

  12. Viktor Sanca and Anastasia Ailamaki. 2022. Sampling-Based AQP in Modern Analytical Engines. In Data Management on New Hardware. Association for Computing Machinery, New York, NY, USA, Article 4, 1–8. Link - Presented at DaMoN@SIGMOD’22

  13. Panagiotis Sioulas, Viktor Sanca, Ioannis Mytilinis, and Anastasia Ailamaki. 2021. Accelerating Complex Analytics using Speculation. In Conference on Innovative Data Systems Research. Link - Presented at CIDR’21