Remote job
Senior Data Scientist
Job details
About this role
Role overview
Help take a product knowledge graph from research prototype to production for a large e-commerce catalog. The work combines applied machine learning, language understanding, graph modeling, and data engineering to capture how products relate, what they are used for, and what shoppers may intend to find.
Responsibilities
- Design, train, and evaluate language or graph-based models that identify product concepts, build hierarchies, and connect concepts to catalog items at scale. - Develop data pipelines and generate product and concept embeddings for downstream uses such as recommendations. - Run research and experiments: review relevant work, define evaluation measures, refine prompts, and improve annotated datasets. - Prepare models and data products for production with engineering partners, including support for controlled A/B evaluations. - Build and assess semantic relationships that go beyond product titles and standard attributes, such as complementarity and usage context.
Requirements
- Around four to five or more years of commercial experience in data science or machine learning research on large-scale projects. - A university degree in computer science, mathematics, statistics, physics, or a related quantitative field. - Experience taking R&D from literature review and hypotheses through experimental validation to production-ready delivery. - Practical NLP and LLM experience, including prompt design, information extraction, synthetic data, semantic clustering, Transformer models, and sentence encoders. - Strong Python and machine learning skills, with tools such as PyTorch or TensorFlow, Hugging Face Transformers, Pandas, NumPy, and scikit-learn; ability to write maintainable, testable code and work with CI/CD. - Advanced SQL and experience processing large datasets in cloud environments. The source also mentions recommendation and semantic search systems, data annotation operations, and Google Cloud tools, though the relevant requirements text is truncated.
Nice to have
- Experience with recommendation architectures such as two-tower retrieval and ranking, complementary-item recommendations, or approximate nearest-neighbor/vector search. - Experience creating annotation guidelines, coordinating annotators, and assessing label quality and consistency.