Overview
In this role you will build end-to-end ML models for retrieval, ranking, and personalization to help buyers find the perfect items in Vinted’s vast catalog. You will work closely with engineers to deploy models in high-throughput, low-latency search and recommendation systems. You’ll design evaluations and experiments to quantify business impact and collaborate with product and data partners to translate problems into ML solutions. This is a chance to shape discovery at scale in a collaborative, research-driven environment with an in-house experimentation platform.
Arbeidsvoorwaarden / Benefits- share options programme
- 25 days of paid annual leave
- Newest MacBook models
- Lunch benefit
- Learning budget
- Hybrid work
Verantwoordelijkheden- Build and optimize end-to-end machine learning models for retrieval, ranking, and personalization
- Collaborate with platform and software engineers to serve models in search and recommendation engines with millisecond latency
- Lead or contribute to design proposals for new modeling approaches, features, or training pipelines
- Define tracking requirements, perform offline evaluations, and design online A/B tests to measure business impact
- Collaborate with Data, Engineering and Product stakeholders to identify opportunities and align on technical trade-offs
- Share knowledge and contribute to the ML Guild and Data Science & Analytics growth
Belangrijkste vereisten- Proficiency in Python and SQL
- Experience with deep learning frameworks and gradient boosting libraries
- Solid understanding of information retrieval, Learning-to-Rank, semantic search, or sequence-based modeling
- Experience or interest in high-performance model serving
- Familiarity with Vespa or Elasticsearch, and serving frameworks like Triton is a plus
- Experience with cloud-based data warehouses and large datasets
- Strong grasp of statistics, offline metrics, and online A/B testing
- collaborative
- strong communication
- problem-solving mindset
- Python
- SQL
- deep learning frameworks
Solliciteren