Looking for a job

Felipe San Martin

@felipesanmartin

Data Scientist at MATRIX CONSULTING who cut an AI model’s runtime from 4 hours to 20 minutes for a multinational client.

Chile

Message

At MATRIX CONSULTING, I technically led an AI product for forecasting, recommendations, and prediction for a multinational client. I developed XGBoost and CatBoost models that achieved over 88% accuracy and are applied in the client's operations across multiple US states.

I redesigned the product architecture on AWS and Azure, reducing model runtime from 4 hours to 20 minutes. My engineering thesis modeled recyclable-waste collection routing and implemented a Max-Min Ant System that found routes up to 5.1% shorter than Gurobi on the same instances.

Experience

Work history, roles, and key accomplishments

Semi-Senior Data Scientist

Matrix Consulting

Jul 2026 - Aug 2026 (1 month)

Technically led an AI product for forecasting, recommendation and prediction for a multinational client, setting objectives with management and overseeing delivery within a team of 5. Developed predictive models achieving over 88% accuracy and redesigned architecture on AWS and Azure, cutting model runtime by 92%.

XGBoost CatBoost AWS Sagemaker AWS ECS Lambda S3 Cloudwatch Azure GitHub Technical Leadership Forecasting Recommendation Systems predictive modeling

Junior Data Scientist

Matrix Consulting

Dec 2024 - Jul 2026 (1 year 7 months)

Developed predictive models (XGBoost, CatBoost) achieving over 88% accuracy validated through field visits, now applied by the client in operations across multiple US states. Redesigned architecture on AWS and Azure, cutting model runtime from 4 hours to 20 minutes, and introduced version control and technical documentation.