BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//EURO Practitioners' Forum - ECPv6.15.13.1//NONSGML v1.0//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-ORIGINAL-URL:https://www.euro-online.org/websites/or-in-practice
X-WR-CALDESC:Events for EURO Practitioners' Forum
REFRESH-INTERVAL;VALUE=DURATION:PT1H
X-Robots-Tag:noindex
X-PUBLISHED-TTL:PT1H
BEGIN:VTIMEZONE
TZID:UTC
BEGIN:STANDARD
TZOFFSETFROM:+0000
TZOFFSETTO:+0000
TZNAME:UTC
DTSTART:20250101T000000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTART;TZID=UTC:20260410T100000
DTEND;TZID=UTC:20260410T110000
DTSTAMP:20260414T194526
CREATED:20260313T142357Z
LAST-MODIFIED:20260313T142357Z
UID:2196-1775815200-1775818800@www.euro-online.org
SUMMARY:A practical approach to replenishment optimization with extended (R\, s\, Q) policy and probabilistic models
DESCRIPTION:Speakers: Alva Presbitero (Senior Applied Scientist\, Zalando) and Shikhar Dev (Senior Machine Learning Engineer\, Zalando) \nIn the high-stakes world of e-commerce\, inventory management is a constant “Inventory Paradox”: carry too much stock and capital is trapped in liquidation; carry too little and you face the “silent killer” of retail—stock-outs. This webinar demonstrates how the ZEOS Inventory Optimization Tool addresses these challenges by unifying probabilistic demand forecasting with Monte Carlo discrete event simulation to drive optimal replenishment decisions. \nThe session is divided into two specialized deep dives: \nPart 1: Applied Science Deep Dive \nWe explore the research and methodology behind the engine\, focusing on these key technical learnings: \n● Extending classical (R\, s\, Q) policies: How we created recommendations that actually fit the fast-paced e-commerce landscape. \n● Probabilistic Demand Modeling: Why our backtest results suggest that modeling full demand probability distributions is far more effective than single-value forecasts. \n● Cost Profile Optimization: How combining probabilistic forecasting with the distribution of our cost profile (optimizing for the 75th percentile) yielded significantly better results than simply looking at the mean. \nPart 2: Machine Learning Engineering Deep Dive \nWe reveal the technical architecture and tools used to put the system into production at scale: \n● Monte Carlo Simulation at Scale: How we used discrete event simulation to stress-test policies across thousands of “alternate timelines” to find the optimal replenishment timing and quantities. \n● Modern ML Stack: A look at how we transitioned research code to scalable and efficient pipelines using AWS\, SageMaker\, and Databricks\, orchestrated by zFlow. \n● Scaling to Millions: Managing feature engineering\, online feature stores\, and real-time inference for millions of SKUs. \nReferences: \n1. AS: https://www.nature.com/articles/s41598-025-32537-2 \n2. MLE: https://engineering.zalando.com/posts/2025/06/inventory-optimisation-system.html \n  \nAbout the speakers: \nAlva Presbitero\, Senior Applied Scientist\, Zalando \nAlva is a Senior Applied Scientist at Zalando\, focusing on the research and data science that powers the ZEOS replenishment engine. She earned her PhD in Computational Science at the University of Amsterdam\, where her research focused on modeling the human innate immune system. Drawing on an extensive background in applied science\, her earlier work was rooted in complexity science\, agent-based modeling\, and computational immunology. \nToday\, she loves using that foundation in complex\, stochastic systems to build scalable and smart inventory solutions for the world of e-commerce. \nShikhar Dev\, Senior Machine Learning Engineer\, Zalando \nShikhar Dev is a founding member of the data and ML engineering team at ZEOS\, where he established the vision and technical foundations for the department’s ML infrastructure. With over a decade of experience and a Master’s in Computer Science from TU Delft\, Shikhar specializes in bridging the gap between research and production. Beyond his work at ZEOS\, Shikhar contributes to Zalando’s broader ML Engineering strategy while also laying the technical foundations for the next generation of GenAI solutions within the department. \n 
URL:https://www.euro-online.org/websites/or-in-practice/event/a-practical-approach-to-replenishment-optimization-with-extended-r-s-q-policy-and-probabilistic-models/
ATTACH;FMTTYPE=image/png:https://www.euro-online.org/websites/or-in-practice/wp-content/uploads/sites/8/2026/03/2.png
END:VEVENT
END:VCALENDAR