Data Scientist: Forecasting & Ranking for Commerce AI

Topsort; Santiago; On-site; CLP 42,000,000 - 66,000,000; Full time

Job summary

Topsort, a global Commerce AI company, is seeking a Data Scientist to improve ranking and retrieval for search and ads, and to build predictive models for advertiser planning. You will collaborate with Product and Engineering to turn data into actionable insights, ship models into our ad engine, and support multilingual catalogs and large-scale event data pipelines.

We value curiosity, strong judgment, and ownership of results, with a bias toward production-ready solutions that deliver

Qualifications

• 5+ years of data science experience with a focus on forecasting and predictive modeling.
• Experience building search, recommendation or ads ranking systems in production.
• Hands-on experience with PyTorch and embedding-based retrieval/ranking models.

Responsibilities

• Develop and refine forecasting models for budget planning and advertiser spend using time-series methods and ML.
• Build and improve ranking/retrieval models for search and placements.
• Represent queries and products, including multilingual catalogs and cold-start scenarios.
• Explore large datasets to identify trends and improve predictions.
• Ship models into production and build training/evaluation pipelines for large-scale event data.
• Collaborate with Product and Engineering to integrate models into platform workflows.

Skills

Forecasting

Ranking & Retrieval

Python

SQL

Experimentation

MLOps

Education

Ph.D. in Data Science or related field

Tools

PyTorch

FAISS

HNSW

Job description

Topsort, a global Commerce AI company, is seeking a Data Scientist to improve ranking and retrieval for search and ads, and to build predictive models for advertiser planning. You will collaborate with Product and Engineering to turn data into actionable insights, ship models into our ad engine, and support multilingual catalogs and large-scale event data pipelines.

We value curiosity, strong judgment, and ownership of results, with a bias toward production-ready solutions that deliver