Solving Marketplace Cold Start at Scale with Ranking
How to handle supplier, product, and demand cold start in a two-sided marketplace by combining ranking signals, exploration, experimentation, and marketplace-level objectives.
Theodore Meynard
Machine Learning Leader | Ranking, Search & Marketplace Optimization · Berlin, Germany
I am a machine learning leader with 8+ years of experience building ranking, recommendation, search relevance, and optimization systems for large-scale marketplaces. I currently lead Ranking & Search Relevance at GetYourGuide, where my team owns ML systems influencing millions of travelers across a significant majority of a €4B+ GMV marketplace. I am also active and recognized in the European PyData community as an organizer and conference speaker.
Experience
Feb 2023 – Present
Data Science Manager, Ranking & Search
GetYourGuide Germany
- Lead a cross-functional ML team of 10+ people, including 6 direct reports across Data Science and MLOps; doubled the team size, hired 4 people, and promoted 4 team members to Senior.
- Own ranking and search relevance systems affecting millions of travelers and influencing a significant majority of GetYourGuide's €4B+ GMV marketplace.
- Led the evolution of ranking from static scoring to GBDT and state-of-the-art deep learning ranking architectures (ranking journey, deep learning migration).
- Conceived and secured executive support for a multi-objective ranking strategy balancing short-term revenue, customer value, and long-term marketplace health.
- Created and now lead a cross-functional Search Relevance virtual team focused on hybrid retrieval, personalized query suggestions, and LLM-based query understanding.
- Built operational and engineering practices that became a reference for other ML teams.
- Led a Claude-powered ML system health and observability initiative adopted across the ML department.
Aug 2020 – Feb 2023
Senior / Staff Data Scientist, Ranking & Recommendations
GetYourGuide Germany
- Drove multiple percentage points of global Net Revenue uplift through relevance modeling, ranking improvements, and dozens of online A/B experiments.
- Defined and executed the technical vision for ranking as an independent service, creating the foundation for a dedicated Ranking team.
- Built the first internal ML platform adopted by 30+ practitioners across the ML department, standardizing batch and real-time inference with PySpark, Airflow, FastAPI, and MLflow (platform foundations, real-time inference).
Oct 2018 – Jun 2020
ML Engineer, Recommendations & Platform
plista (WPP) Germany
- Improved recommendation CTR by around 20% using XGBoost ranking models and BERT-based semantic features for multilingual recommendations.
- Migrated the entire recommendation stack to AWS, including models, Cassandra-based feature storage, and streaming pipelines.
- Operated recommendation systems serving hundreds of millions of impressions per day.
Sep 2017 – Oct 2018
Data Scientist
optilyz Germany
- First and sole data scientist at an early-stage startup.
- Built data quality and fraud detection systems for marketing campaigns and improved campaign ROI through postcode-level targeting analysis.
Selected Talks
All talksFrom Trees To Transformers: Adopting Deep Learning for Ranking
Designing, training, and evaluating deep learning models for marketplace ranking systems, with real-world experimentation insights.
Switching from Data Scientist to Manager
Reflections and practical advice on transitioning from an individual contributor role in data science to an engineering management position.
Data Unit Tests
Applying software engineering testing principles to data pipelines and machine learning systems to improve reliability and reproducibility.
ML Platform Architecture
Building and scaling an internal ML platform to support training, deployment, and experimentation workflows across data teams.
Community
EventsPyData Berlin steering committee member since 2019. I help organize conferences for 500-1,500 attendees, coordinate 15-20 organizers and around 50 volunteers, and organize monthly meetups for around 100 data and ML practitioners.
Education
2017
Final-Year Internship — CIMEX Hardware Adaptation
Airbus Defence and SpaceGermany
Six-month internship assessing how CIMEX, an ESA microgravity experiment studying heat and mass transfer at liquid interfaces, could be adapted to an alternative uncrewed spacecraft after its planned ISS deployment was discontinued. Translated scientific objectives into technical requirements, investigated the necessary hardware modifications, and consulted specialists to define a viable solution within cost and schedule constraints.
2017
Engineer's Degree — Aeronautics Engineering (Double Degree)
University of São PauloBrazil
Master thesis in image processing and computer vision: measuring flexible wing displacement using camera-based post-processing. Focused on experimental validation, numerical simulation, and data-driven modeling. Conducted one-year research project on aerodynamic instability using Direct Numerical Simulation (Fortran, Matlab); published at EPTT 2016 conference (São José dos Campos, Brazil).
2017
Master of Engineering (M.Eng.)
École Centrale ParisFrance
French Grande École engineering programme (2012-2017) with strong foundation in applied mathematics, statistical modeling, optimization, and computational physics. Teaching assistant in mathematics for first-year students.
2015
Engineering Internship — CIMEX Integration & Testing
Airbus Defence and SpaceGermany
Six-month internship integrating, testing, and improving the CIMEX engineering model with an international engineering team during its final integration phase. Gathered technical findings from the engineering model to inform the future development of the flight model intended for the Fluid Science Laboratory aboard the International Space Station.
2014
Research Internship — Fluid Dynamics & Computational Modelling
Cornell UniversityUSA
Research at the Fluid Dynamics Research Laboratories on unsteady propulsion from flexible membranes. Designed experiments, implemented data acquisition pipelines in Python, and performed quantitative analysis of complex physical systems under the supervision of Riley Shutt.
2012
Classe Préparatoire — Advanced Mathematics & Physics
Lycée Pierre de FermatFrance
Intensive two-year programme preparing for French Grandes Écoles, focused on advanced mathematics, problem solving, and theoretical physics.
2010
Baccalauréat Scientifique (S)
Lycée Gustave EiffelFrance
Skills
Marketplace & Optimization
- Ranking
- Recommendations
- Search Relevance
- Learning-to-Rank
- Multi-objective Optimization
- Candidate Selection
- Experimentation
Machine Learning
- Deep Learning
- Prediction Systems
- Relevance Modeling
- Calibration
- A/B Testing
ML Engineering
- ML Platforms
- MLOps
- Real-time Inference
- Batch Inference
- Observability
- Model Deployment
Technologies
- Python
- SQL
- PySpark
- Airflow
- MLflow
- FastAPI
- AWS
- Terraform
Leadership
- Hiring
- Coaching
- Technical Strategy
- Stakeholder Management
- Organizational Design
Community
- PyData Berlin steering committee member since 2019
- Conferences for 500-1,500 attendees
- Coordination of 15-20 organizers and around 50 volunteers
- Monthly meetups for around 100 practitioners