Vandemberg Monteiro
Data Scientist experienced in designing and deploying data-driven solutions across finance and healthcare.
Skilled at translating complex analytics into actionable insights and supporting effective decision-making with robust, impactful solutions.
View My LinkedIn Profile
Selected Projects in Data Science and Machine Learning
End-to-end credit risk modeling with decision and policy analysis
In this project, I develop an end-to-end credit risk modeling pipeline, focusing not only on predictive performance but also on decision-making, model stability, and risk policy evaluation.
The work includes feature engineering and preprocessing, gradient boosting models for risk ranking, model interpretability with SHAP, stability and behavioral analysis over time, and the translation of scores into approval policies and risk–return trade-offs.
The project also explores deployment considerations and causal analysis to better understand drivers of default beyond correlation.
Conversational recommender systems with agentic RAG and ranking models (Ongoing)
In this project, I design and implement a decision-aware recommender system that combines classical recommendation models with a conversational LLM agent grounded via retrieval-augmented generation (RAG). Rather than allowing the LLM to directly select or rank items, the agent is constrained to translate unstructured user preferences into explicit, validated constraints and ranking boosts, ensuring controllability and auditability.
The system compares multiple recommendation approaches — from simple popularity baselines to collaborative filtering and state-of-the-art two-stage ranking — and evaluates how conversational signals affect ranking quality, cold-start performance, and recommendation behavior under realistic data constraints. Special emphasis is placed on offline evaluation, error analysis, and failure modes, reflecting production-oriented design choices over purely theoretical gains.