Assistant Director, Data Science

Role Title: Assistant Director, Data Science

Department: Data Science Excellence, US Data Science – USRM
Location: Colombia
Type: Full Time
Reporting to: Modeling Sophistication Manager, Pedro Silva

About the Role:
Are you curious, resilient, independent, and always interested in learning and working on state-of-the-art modeling techniques, applying the best science to deliver business value? The USDS Data Science Excellence team is looking for an Assistant Director to research, design, and build major modeling innovations and pioneer model sophistication across our retail insurance products.

Key Responsibilities:
• Design, develop, and scale gradient-boosted decision-tree models end-to-end: problem framing, data wrangling, exploratory data analysis (EDA), feature engineering, hyperparameter optimization, and model evaluation.
• Build and optimize large-scale training workflows (e.g., parallel/distributed training on AWS EMR or similar environments).
• Research and experiment with new modeling techniques for structured and unstructured data, and evaluate emerging ML tools and best practices.
• Own pioneering projects from research through proof of value and product ownership; identify and remove obstacles to deployment in partnership with infrastructure and design teams.
• Research, design, and assist with the implementation of models in Earnix and other platforms.
• Adopt a transformational mindset to improve and automate processes where applicable.
• Regularly engage with the data science community and participate in cross-functional working groups.

About US Data Science:
USDS employs more than 350 full-time professionals who build advanced data science products and AI solutions for our US Retail Markets (USRM) business. By thoughtfully applying data science across USRM functions, we ensure we consistently deliver on our promises while driving sustainable growth across our portfolio. We accomplish this through a focus on model development, deployment, and decision science. In model development, we design sophisticated models that combine deep data science expertise with close collaboration across our go-to-market verticals. This integration ensures our models not only incorporate cutting-edge science but, more importantly, generate measurable business value aligned with each vertical’s strategic goals.

Skills and Experience:
• Experience building gradient-boosted decision trees and performing model interrogation.
• Strong coding experience; fluent in object-oriented programming, Python, and common Python data and modeling packages (LightGBM, Optuna, SHAP).
• Good software practices: version control (Git), pre-commits, code reviews, documentation.
• Experience with EDA and data preparation for building gradient-boosted decision trees.
• Knowledge of Earnix or similar rating tools is a plus.
• Demonstrated ability to exchange ideas and convey complex information clearly and concisely.
• A value-driven perspective on work context and impact.
• Demonstrated professional-level proficiency in English, with excellent written and verbal communication skills.
• Education and experience requirements:
o PhD (in a quantitative field) + 2+ years relevant experience, or
o Master’s + 3+ years relevant experience, or
o Bachelor’s + 5+ years relevant experience.

About Liberty Mutual:
As a purpose-driven organization, Liberty Mutual is committed to fostering an environment where employees from all backgrounds can build long and meaningful careers. Through strong relationships and continuous learning opportunities, we seek to create an environment where employees can succeed, both professionally and personally.

At Liberty Mutual, we believe progress happens when people feel secure. By providing protection for the unexpected and delivering it with care, we help people embrace today and confidently pursue tomorrow.