SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Topsort is the leading Retail Media Infrastructure company, democratizing auction-based ad technology for retailers, marketplaces, and delivery apps across 40+ countries. We're a fast-growing, global-first team of ~100 people with offices in Boston, Santiago, São Paulo, Barcelona, Sydney, and beyond.
You'll join as a Data Scientist to improve how Topsort ranks and retrieves products for search and ads, predicts user behavior, and supports advertiser planning through reliable predictive models. Working closely with Product and Engineering, you'll turn complex data into actionable models and insights that drive better decisions and continuously improve our platform.
Key responsibilities include:
- Developing and refining forecasting models for budget planning and advertiser spend using time-series methods, regression, and machine learning
- Building and improving retrieval models that select which products are candidates for each search or placement
- Developing and iterating on rankers and re-rankers that combine relevance, CTR/CVR predictions, and business constraints for organic and sponsored placements
- Improving query and product representation, including multilingual catalogs, sparse or noisy metadata, and cold-start scenarios
- Exploring large datasets to identify trends, assess data quality, and uncover opportunities to improve predictions
- Shipping models into the ad engine and serving stack, building training and evaluation pipelines on large-scale event data
- Partnering with Product and Engineering to integrate models into platform workflows and translate findings into actionable recommendations
Success means delivering reliable predictions with consistent performance against baselines, producing measurable lift in ranking and retrieval for shoppers and advertisers, providing useful budget forecasts for planning, ensuring practical adoption by product and engineering teams, and driving continuous improvement based on model errors and experiment results.
Topsort's culture emphasizes straightforward communication, embracing feedback, collaboration, working in the open, and maintaining a high-velocity pace. The team values ownership, production-grade code, honest tradeoffs, attention to detail (especially in infrastructure), clear communication, and intellectual curiosity.
**Requirements:**
- 5+ years of experience in data science with strong focus on forecasting
- Ph.D. in Data Science, Statistics, Mathematics, or related quantitative field
- Direct experience building search, recommendation, or ads ranking systems in production
- Hands-on experience with PyTorch or similar, including training embedding-based retrieval or ranking models
- Strong Python and SQL; comfortable building datasets from raw event logs at scale
- Solid statistics fundamentals and experience designing and reading A/B tests
- Comfort with AI-assisted development tools and bias toward getting models into production
**Nice-to-have:**
- Retail media / AdTech experience, auction or bidding systems
- LLM-based retrieval, model distillation, approximate nearest neighbor search (HNSW, FAISS)
- Learning-to-rank at scale or multi-objective ranking
