end-to-end data science lifecycle)  2012-2018 IBM Research – Almaden, USA  Declarative large-scale machine learning  Optimizer and runtime of Apache SystemML  2011 PhD TU Dresden, Germany  Cost-based Database Systems (ADBS, WS) Architecture of ML Systems (AMLS, SS) Data Integration and Large-Scale Analysis (DIA, WS) Master Bachelor Data management from user/application perspective Distributed Boehm, Peter J. Haas, Frederick R. Reiss, Berthold Reinwald: Compressed Linear Algebra for Large-Scale Machine Learning. PVLDB 2016] [Ahmed Elgohary, Matthias Boehm, Peter J. Haas, Frederick R. Reiss

Execution Time [ms])  Don’t cheat or mislead readers and reviewers  Start y-axis at 0 for linear scale Experiments and Result Presentation Runtime What are the units? Where are the tics? Runtime range to make you look good If there are multiple relevant parameters, show them all For log-scale you can’t start at 0 [J. Sommer, M. Boehm, A. V. Evfimievski, B. Reinwald, P. J. Haas: MNC: Structure- Presentation – Result Interpretation [Matthias Boehm et al: On Optimizing Operator Fusion Plans for Large-Scale Machine Learning in SystemML. PVLDB 11(12) 2018] 19 706.015 Introduction to Scientific Writing

understands it, such understanding is not guaranteed to hold a few years from now. Be mindful of our scale With a codebase of 100+ million lines and thou- sands of engineers, some mistakes and simplifications become visible when the user does something wrong. Template metaprogramming interferes with large scale refactoring by making the job of refactoring tools harder. First, the template code is expanded in