¡Hola! I am a Machine Learning Engineer and Data Scientist with 5+ years across industry and research, currently completing an MSc in Mathematical Engineering (Statistical Learning track) at Politecnico di Milano on an Invest Your Talent in Italy government scholarship. Alongside my studies I work as an AI Engineer at IMHTP in Milan, designing multi-agent LLM pipelines and real-time computer-vision systems.
Before moving to Italy I was a Semi-Senior AI / Data Science Engineer at Mercado Libre, where I shipped probabilistic forecasting for inventory optimization and pioneered reinforcement-learning and contextual-bandit models for dynamic pricing. Earlier, I worked as a data-science consultant at Quantil, combining applied research with client projects, and as a research assistant at Universidad del Rosario's Tic Tank group, which led to a workshop paper at ICML 2022 (LatinX in AI).
My background bridges economics, applied mathematics, and software engineering. I am especially interested in sequential decision-making, agentic LLM systems, 3D computer vision, and making things fast on GPUs.
2025Published an interactive blog on robust superquadric fitting for messy point clouds (STAG 2025 material).
2025Joined IMHTP in Milan as AI Engineer / Data Scientist while finishing my MSc.
2025Completed NVIDIA's Accelerated Computing in CUDA C/C++ certification.
2024Started the MSc in Mathematical Engineering at Politecnico di Milano with the Invest Your Talent in Italy scholarship.
2022 · 07Presented A Model-Based Filter to Improve Local Differential Privacy at the LatinX in AI workshop, ICML 2022 (Baltimore).
↘ 02
Experience
2025 – now
AI Engineer / Data Scientist
IMHTP · Milan, Italy
Multi-agent LLM pipelines (foundation models + NLP pattern matching) for automated healthcare data processing, reaching error-free extraction on all manually validated cases. Prototyped STT/TTS models and a real-time computer-vision anomaly-detection pipeline (DRAEM + temporal deep learning, 91% accuracy) for industrial process evaluation. Benchmarked survival-analysis models (Cox PH, Kaplan–Meier, XGBoost Survival, DeepSurv) to a 0.70 concordance index on clinical data.
2022 – 2024
Semi-Senior AI / Data Science Engineer
Mercado Libre · Bogotá, Colombia
Improved median stock health by ~15% and cut overstock by 10% with probabilistic forecasting integrated into inventory optimization. Pioneered reinforcement learning and contextual bandits for dynamic pricing and order-quantity models. Built LLM agents with multi-agent workflows and Tree-of-Thought prompting for executive decision support; contributed to a Stable Diffusion image-generation initiative and a Tableau seller-funnel dashboard.
2022 – 2024
Data Science Code Reviewer
TripleTen LatAm · Remote
Reviewed and validated 200+ student projects across Python, SQL, statistics, supervised/unsupervised ML, time series, NLP, and computer vision — mentoring learners on accuracy and best practices.
2021 – 2022
Data Scientist (Consulting & Research)
Quantil · Bogotá, Colombia
Built a record-matching algorithm achieving an 80% match rate on partial data. Implemented gas-shortage estimation methods that reduced client costs by 25%. Time series, supervised learning, and dynamic panel-data models in Python, R, and STATA.
2020 – 2021
Research Assistant
Universidad del Rosario · Tic Tank research group
Researched differential-privacy algorithms, contributing to a published workshop paper at ICML 2022 (LatinX in AI).
Extended GAIR-RANSAC to decompose noisy point clouds into superquadric primitives, cutting average Chamfer distance by 40% across synthetic scenes and real Thingi10K meshes. Drag the 3D figures to explore.
Presented in Baltimore, MD. Proposes a model-based filtering step that improves the utility of locally differentially private data releases while preserving the privacy guarantee.
Certifications
2025Accelerated Computing in CUDA C/C++ — NVIDIA
2023Deep Reinforcement Learning (Stable Baselines3, Gymnasium) — Hugging Face
