Open to opportunities

Faber Santiago Herrera Sanguino

@fabersantiagoherrera

Message

Data Science Consultant at GMS Management Solutions SAS who lifted expected profit ~6% with a price-optimization model.

Colombia

Message

At GMS Management Solutions SAS, I built clustering models to segment more than 1 million insurance policies and explain IFRS 17 reserve drivers to actuarial and finance stakeholders. I also developed a price-optimization model that lifted portfolio expected profit by about 6%.

I designed an AI agent to streamline underwriting and automated data pipelines for an IFRS 17 reporting transition. In research, I built Bayesian parameter-estimation pipelines at Universidad de Los Andes that reduced runtime from about two weeks to about two hours, and developed signal-processing algorithms at IJCLab, Université Paris-Saclay.

Experience

Work history, roles, and key accomplishments

Current

Data Science Consultant

Current

GMS Management Solutions SAS

Jul 2025 - Present (1 year 4 months)

Built clustering models to segment 1M+ insurance policies into homogeneous risk groups and developed a price-optimization model that lifted portfolio expected profit by ~6%. Automated end-to-end data pipelines for IFRS 17 regulatory reporting and worked directly with client actuarial, finance, and accounting teams.

Python SQL R Clustering Segmentation Classification XGBoost Random Forest Feature Engineering Model Validation Price Optimization IFRS 17 Data Pipelines AI Agents Stakeholder Management

Current

AI Evaluation & Prompt Engineering Contributor

Current

Outlier

Dec 2024 - Present (1 year 11 months)

Stress-tested LLM robustness across reasoning, code-generation, and multimodal tasks through adversarial prompt design and structured benchmarking, surfacing systematic failure modes.

LLM Evaluation Prompt Engineering Adversarial Testing Benchmarking

Research Intern

IJCLab, Université Paris-Saclay

May 2025 - Jul 2025 (2 months)

Developed a modular Python/C++ signal-processing algorithm achieving 95% reconstruction accuracy on high-noise detector signals. Designed a reproducible data-analysis pipeline ensuring data integrity across ~1 GB/hour of experimental acquisition data.

Python C++Signal Processing Clustering Filtering Data Pipelines

Research Assistant

Universidad de Los Andes

Aug 2024 - Dec 2024 (4 months)

Built Bayesian parameter-estimation pipelines (MCMC) for high-dimensional models, reducing end-to-end runtime from ~2 weeks to ~2 hours. Automated computational workflows in Python and validated results against experimental data.

Python Bayesian Inference (PyMC)MCMC Statistical Inference Data Pipelines Automation