Data Scientist (Pre-Search)

Responsibilities

• Develop and improve query understanding capabilities for Trendyol's search experience across multiple regions and languages, including spelling correction, query normalization, query rewriting, intent detection, and semantic retrieval.
• Build and evaluate vector search, embedding, and retrieval models that improve the relevance of product discovery for customers across different markets.
• Take a leading role in Arabic-language search initiatives, addressing language-specific challenges such as dialectal variation, orthographic normalization, transliteration, morphology, mixed-language queries, and spelling errors.
• Analyze large-scale structured and unstructured data, including search queries, product content, user interactions, and linguistic signals across the Trendyol ecosystem.
• Design robust offline evaluation methodologies and quality metrics for query understanding, spelling correction, and retrieval systems; validate improvements through online A/B experiments.
• Collaborate closely with Search, Pre-Search, Ranking, Product, and Engineering teams to translate customer and business needs into scalable machine learning solutions.
• Improve model quality, latency, reliability, and deployment cycles by applying MLOps best practices to production ML systems.
• Monitor model and search-quality performance after deployment, investigate regressions, and continuously iterate based on data and user feedback.
• Communicate methodology, findings, and expected business impact clearly to technical and non-technical stakeholders.

Expected Qualifications

• Experience in applied data science and machine learning, ideally in NLP, information retrieval, search, or language modeling domains.
• Professional working proficiency in Arabic, with a strong understanding of Arabic grammar, morphology, spelling conventions, dialectal variation, and common search-query behavior. Native-level proficiency is a strong plus.
• Practical experience with modern NLP methods, including transformer-based language models, text embeddings, semantic search, retrieval, reranking, query understanding, or spelling correction.
• Familiarity with Arabic NLP resources, datasets, tokenization or normalization approaches, and evaluation challenges is highly preferred.
• Strong Python skills and hands-on experience with relevant ML libraries such as PyTorch, TensorFlow, scikit-learn, Hugging Face Transformers, or similar frameworks.
• Strong SQL knowledge and familiarity with data-analysis best practices.
• Experience with vector databases, approximate nearest neighbor search, embedding models, or large-scale retrieval systems is a strong plus.
• Experience designing and interpreting A/B experiments, offline evaluations, and statistical analyses.
• Strong analytical skills, a data-driven mindset, and the ability to work effectively with ambiguous product problems.
• Excellent written and verbal communication skills in English. Turkish is a plus.
• Experience in e-commerce, marketplaces, search, recommendation, or self-service platforms is a plus.