Alexander Renz-Wieland
Computer Scientist
I enjoy developing complex software that—hopefully—makes the world a tiny bit better. Currently, I try to do that with a fantastic team of people at RelationalAI.
Prior, I developed systems for training large and/or complex machine learning models efficiently on computer clusters, working as a PhD candidate in the DIMA research group at Technische Universität Berlin, supervised by Volker Markl and Rainer Gemulla. More details in my thesis, several papers on the topic, and this interactive visualization of the basics.
And before that, I studied computer science and business, with a focus on large-scale data management, at Universität Mannheim and VU Amsterdam / CWI, and wrote a master's thesis on scalable sequential pattern mining. Even prior, I studied in a dual program, working for HP in Germany and the U.S., and studying at DHBW Mannheim and Universidad Carlos III Madrid.
Publications
Full Research Papers
Good Intentions: Adaptive Parameter Management via Intent SignalingA. Renz-Wieland, A. Kieslinger, R. Gericke, R. Gemulla, Z. Kaoudi, V. Markl. CIKM '23.
pdf. arXiv. source code.
NuPS: A Parameter Server for Machine Learning with Non-Uniform Parameter Access
A. Renz-Wieland, R. Gemulla, Z. Kaoudi, V. Markl. SIGMOD '22.
pdf. arXiv. slides. recording. source code.
Dynamic Parameter Allocation in Parameter Servers
A. Renz-Wieland, R. Gemulla, S. Zeuch, V. Markl. PVLDB, 13(11): 1877-1890, 2020.
pdf. arXiv. slides. source code. interactive demo.
Scalable Frequent Sequence Mining With Flexible Subsequence Constraints
A. Renz-Wieland, M. Bertsch, R. Gemulla. ICDE '19.
pdf. slides. poster. source code.
Misc
Adaptive Parameter ServersA. Renz-Wieland. Ph.D. thesis, Technische Universität Berlin, December 2022.
library. pdf. slides. source code.
Just Move It! Dynamic Parameter Allocation in Action
A. Renz-Wieland, T. Drobisch, R. Gemulla, Z. Kaoudi, V. Markl. PVLDB, 14(12): 2707-2710, 2021 (demo).
pdf. website.
The DESQ Framework for Declarative and Scalable Frequent Sequence Mining
K. Beedkar, R. Gemulla, A. Renz-Wieland. INFORMATIK '19 (extended abstract).
pdf. slides. source code.
Distributed frequent sequence mining with declarative subsequence constraints
A. Renz-Wieland. Master's Thesis, Universität Mannheim, April 2017.
pdf. slides. source code.
Airport quality: holding and go-arounds
A. Renz-Wieland, H. Wallenburg. Project report for the 2016 LSDE course.
pdf. website.
Awards
I was a recipient of a Software Campus grant for my PhD research. I.e., the BMBF provided funding for a team of assistant researchers, hardware, and other things.
My M.Sc. faculty honored my master's degree (best 2016/17 degree) and master's thesis (best 2016/17 thesis)
Teaching
Courses
- Information Systems and Data Analysis: >1000 students databases 101 course (summer 2019, summer 2020, summer 2021)
- Database Laboratory: database and SQL fundamentals with many practical examples and exercises (winter 2017/18, summer 2018, winter 2018/19, winter 2019/20, winter 2020/21, summer 2022, winter 2022/23)
- Database Seminar (summer 2021)
- Big Data Analytics Project (winter 2021/22)
Thesis supervision
- Integration of Hardware-Accelerators for an Adaptive Parameter Server
- Adaptive Parameter Management in Machine Learning Systems
- Analysing Dynamic Parameter Allocation in Distributed Machine Learning
- Dynamic parallelism in large-scale graph processing
- Lock-Free Data Structures for Locality-Aware Parameter Servers
- Benchmarking Probabilistic Programming Languages
Climate impact
Different aspects of research activity contribute to the climate crisis. One is long-distance travel. Thus, I try to limit travel.
Like others, I think it is reasonable that academics report the climate impact of their work. As a step in this direction, please find below a list of flights I have taken during my time in academia.
| Year | Flight | Reason | CO2 (kg) |
|---|---|---|---|
| 2019 | Berlin-Beijing, Macau-Beijing, Beijing-Berlin |
Talk and poster presentation at ICDE, Talk at ECNU Shanghai |
4,259 |
| 2017 | Berlin-Munich, Munich-Berlin |
Meetings with industry partners | 216 |
The CO2 emission estimates are taken from atmosfair. If you have tips on how to estimate other emissions of research (e.g., computer and cluster usage, office construction and heating), please let me know.
