Calvin Seward

Johannes Kepler Universität Linz

Doctorate Studies , Machine Learning

January 1, 2016 – January 1, 2019

Humboldt-Universität zu Berlin

Master's degree, Mathematics and Computer Science

January 1, 2008 – January 1, 2012

Colorado State University

Master's degree, Mathematics

January 1, 2005 – January 1, 2006

HMS Analytical Software | Consulting & End-to-End Solutions for Data Science & Analytics

Senior Data Scientist

September 1, 2024 – Present

Berlin, Berlin, Germany

Bolt

Senior Data Scientist

November 1, 2022 – September 1, 2024

Berlin, Germany · Hybrid

Zalando

Applied Scientist with focus on Computer Vision and Reinforcement Learning

January 1, 2020 – November 1, 2022

Zalando

Research Scientist PhD with Focus on Deep Learning

September 1, 2016 – December 1, 2019

Zalando

Data Scientist with Focus on Warehouse Logistics

May 1, 2012 – September 1, 2016

Geomecon

Student Mathematician

April 1, 2011 – April 1, 2012

Potsdam Area, Germany

Invited Speaker at M3 Conference in Cologne -- AI successfully applied

April 1, 2018 – April 1, 2018

Rome wasn't built in a day. The same holds for Artificial Intelligence in your company. Every big player in the Artificial Intelligence space first optimized their core business before applying acquired knowledge to generate new AI powered business models. The amazing self driving cars and machine powered translations built upon successes of lane departure warnings and auto-complete. Starting with an example from warehouse logistics, we'll show how Zalando Research streamlined a core business process while building AI trust and knowledge in the company. Further, we discuss management strategies to secure the success of AI projects. Check out the slides at https://www.slideshare.net/CalvinSeward/ai-successfully-applied

Invited Speaker at AI Summit Vienna -- Deep Learning -- More than Classification

September 1, 2017 – September 1, 2017

Many of Deep learning's first big breakthroughs were in the field of classification, for example recognizing hand written digits or imagenet images. While such results were impressive, they are only a drop in the sea of possible deep learning applications, a sea who's extent we are only just now beginning to discover. In this talk, I'll demonstrate a sampling of and exciting deep learning applications we're working on at Zalando in cooperation with Prof. Dr. Sepp Hochreiter's team at JKU Linz. These include image generation with Generative Adversarial Networks, Semi Supervised Semantic Segmentation, Warehouse Optimization and Recommender Systems. I'll then wrap up with a few technical tips and tricks on how you too can start learning your own deep neural networks. Video: https://www.youtube.com/watch?v=oHve-AX8bPI Slides: https://mostly.ai/summit/slides/Zalando%20Calvin%20Seward.pdf

Invited Speaker at data2day Heidelberg -- Search at Petabyte Scale

September 1, 2017 – September 1, 2017

In vielen Big-Data-Anwendungen müssen riesige Datensätze schnell durchsucht werden, um relevante Information wie Kundenprofile, Bilder oder Dokumente zu finden. Der naive Suchaufwand wächst linear mit der Zahl der gespeicherten Daten: ein tödliches Problem für skalierbare Real-Time-Big-Data-Lösungen. Approximate-Nearest-Neighbor-Methoden (ANN) finden die gesuchten Ergebnisse mit hoher Wahrscheinlichkeit, bei zugleich exponenziell reduziertem Aufwand. Diese innovative Technologie ermöglicht erst die Skalierung datenintensiver Anwendungen in den Petabyte-Bereich. In meinem Talk werde ich die Basics von ANN erläutern und eine Bildersuche mittels einer Python-Open-Source-Toolbox demonstrieren.

Invited Speaker at TDWI Munich

June 1, 2017 – June 1, 2017

Title: Leveraging our Pictures: Deep Learning with Tensorflow at Zalando This business oriented talk motivates why data science and machine learning, especially machine learning on unstructured data such as images and natural languages is important for products to be competitive in future. The talk then shows some excited work we've done with images at Zalando research and ends with a quick demonstration of Tensorflow, a framework for easy GPU computing at scale. Click the link to get the slides.

Invited Speaker at Deep Learning in Retail & Advertising Summit London

June 1, 2017 – Present

Title: How Zalando Accelerates Warehouse Operations with Neural Networks This talk introduces a specific problem in the warehouse, the so called order batching problem, and demonstrates how we are able to find near optimal solutions by using machine learning tricks. It's interesting since we're able to apply neural networks outside of the domains they are generally used. The talk concludes with a point that is near and dear to me: only using machine learning to drive efficiency of existing business models leaves many game changing machine learning untapped. Click on link to view the talk. Bonus: there's a video!! watch it before it's pay-walled http://videos.re-work.co/videos/473-deep-learning-for-retail-warehouse-operations

Invited Speaker at Data2Day Karlsruhe

October 1, 2016 – October 1, 2016

Title: Leveraging our Pictures Deep Learning with Tensorflow at Zalando This talk shows an exciting application of Zalando's image data we've been working on at Zalando research which enables us to localize items in a picture without any prior localization training information. The talk concludes with a quick and practical demonstration of Tensorflow, a framework for easy GPU computing at scale. Click the link to get the slides.

Optimizing Warehouse Operations with Neural Networks on GPUs

March 1, 2015 – October 1, 2015

https://devblogs.nvidia.com/parallelforall/optimizing-warehouse-operations-machine-learning-gpus/ We trained a neural network to predict the length of pick routes through the warehouse. Then enables us to quickly try many different combinations of warehouse locations in order to create optimal batches.

OCaPi -- Optimal Picker Routing with Cart

June 1, 2014 – March 1, 2015

https://tech.zalando.com/blog/defeating-the-travelling-salesman-problem-for-warehouse-logistics/ In order to tell guide pickers optimally through the warehouse, we developed a system that finds the optimal route to walk through the warehouse and how pickers can to best manage their carts, resulting in a significant boost in pick performance.

Startup Engineering -- Online Web Development Class

Coursera

June 24, 2026 – Present