Applied AI Stories: Solving Cold Cases and Powering Product Search
PyData Amsterdam is back with another evening of inspiring talks, food, and networking on Thursday the 26th of March, from 17:30 until 21:30 CET at ML6 in Amsterdam! Join us for an exciting meetup where we explore how modern machine learning systems are being applied to real-world problems at scale - from cold case investigations to large-scale e-commerce search. We’ll kick off with drinks and networking, followed by two in-depth technical talks from industry practitioners, with plenty of time to connect over food and drinks afterwards.
Schedule
17.30 – 18.00: 🥤 Welcome & Networking
18.00 – 18.15: ML6 / HR introduction
18.15 – 19.00: 🎤 Talk 1 – Can AI agents help solve Europe’s Largest Cold Case? - By Titus Naber
19.00 – 20.00: 🍕 Food & Networking
20.00 – 20.45: 🎤 Talk 2 – From Keywords to Concepts: A Late Interaction Approach to Semantic Product Search on IKEA.com - By Amritpal Singh
20.40: 21.30: 🥤 Drinks & Networking
Talk 1: Can AI agents help solve Europe’s Largest Cold Case?
by Titus Naber
The Olof Palme investigation lasted 34 years and resulted in one of the largest police archives in modern European history. Even reading the digitized portion alone would take a single person years.
So what happens when you give that archive to AI agents?
In this talk, Titus will walk us through the design and deployment of a Deep Research agent system capable of autonomous hypothesis generation, iterative search, and cross-referencing across thousands of police documents. The session will dive into how AI agents can be orchestrated to explore massive, unstructured datasets and support complex investigative workflows.
Bio
Titus Naber is a Machine Learning Engineer at ML6, based in Amsterdam. He holds an MSc in Artificial Intelligence from Delft University of Technology and has experience across machine learning, full-stack development, and autonomous systems. Previously, he worked at Adyen on open-source payment integrations and was involved in autonomous racing through the Formula Student team. Titus combines strong academic foundations with hands-on industry experience to build practical, high-impact ML systems.
Talk 2: From Keywords to Concepts: A Late Interaction Approach to Semantic Product Search on IKEA.com
by Amritpal Singh
Traditional keyword-based search struggles to capture the rich, multi-attribute intent behind customer queries. In this talk, Amritpal will present IKEA’s journey from keyword matching to concept-based semantic search using a late-interaction retrieval model.
Instead of relying on Boolean logic or single-vector embeddings, this approach uses token-level scoring to preserve keyword specificity while capturing fine-grained intent. The system is trained using large-scale synthetic query generation, strong negative mining, and adaptive ranking — all while maintaining low latency at production scale.
Deployed live on IKEA.com across multiple markets and languages, the new system outperformed the existing Boolean search, delivering meaningful gains in relevance, engagement, and commercial impact, especially for long-tail queries.
Bio
Amritpal Singh is a Senior Data Scientist at IKEA Amsterdam with over 10 years of experience in machine learning, computer vision, and NLP. Previously, he was a Staff Data Scientist at Huawei and a Senior Research Engineer at HERE Technologies. He holds an MSc in Embedded Systems from TU Delft and a BTech in Electrical Engineering from Vellore Institute of Technology.
We look forward to welcoming you for an evening of cutting-edge ML, practical insights, and great conversations - see you there!