Viktor Sanca

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

I am a Doctoral Candidate in the Data-Intensive Applications and Systems (DIAS) lab at Doctoral School for Computer and Communication Sciences at EPFL, advised by Professor Anastasia Ailamaki.

My research is in data management systems, specifically high-performance analytical systems on modern hardware, to enable workload-agnostic, timely, interactive, and cost-efficient insights offered by the growing amounts of data. With modern hardware and analytical methods, my interest is equally in the holistic optimization of model-relational hybrid analytics, approximate query processing, and optimizing data management systems for modern hardware platforms.

My goal and vision are to enable truly interactive analytics with zero assumptions about the workload by co-designing the approximation algorithms to be efficient and utilize fully the underlying hardware, exploring the algorithmic bottlenecks, and developing systems and algorithms that maximize work reuse further to reduce the time-to-insight on high-bandwidth, low-latency analytical systems. Finally, the goal is to develop approximate query processing systems that outperform the best and most highly optimized modern analytical engines, including model-driven analytics. To unlock this potential of modern analytics, I am working and developing my work on approximate analytics on top of a high-performance heterogeneous engine Proteus. Through holistic design and optimization, from query planning to hardware, we can enable cost and time-efficient insights by spending less (computational) time in the fastest layers of increasingly powerful computational hardware while using the latest analytical methods. With the provenance of machine learning models enabling unstructured data analysis, making these methods part of an extended relational model and tightly integrated with optimization, execution, and data access is on my ongoing research agenda. I summarize this in the statement that we need to go beyond vector indices only, and provide both performance and functionality for novel workloads and methods in a declarative an easy fashion for the user.

Outside my regular research and teaching activities, I am a committee member of EPFL IC Ph.D. student association, bringing together the faculty, alumni, industry, and student community to exchange ideas and experiences. This provides a doctoral-school level platform to get together and interact with distinguished faculty members more directly and informally and a forum for discussing challenges and opportunities in industry and organizations.

Outside the office, I enjoy traveling, hiking, singing, cooking, and improving my piano playing skills.

You can download my most recent CV by clicking here. Last updated on December 10, 2023