nova-ims

Multiclass ML model to predict Workers' Compensation Claim Injury Types | NOVA IMS 2024/25

10-month sales forecasting for SIEMENS Smart Infrastructure | XGBoost RMSE 15.54% | ARIMA/SARIMA market indicators | CRISP-DM | NOVA IMS 2025

Used-car price prediction with leakage-safe preprocessing, feature selection and ensemble regression.

RAG chatbot for Fidelidade insurance agents | GPT-4o-mini + Azure OpenAI | 6 languages | Gradio UI | CRISP-DM | NOVA IMS 2025

Customer segmentation for Ready, Steady Ride using K-Means (k=4) across weather, behavior and time perspectives - NOVA IMS ML project

Business Process Optimization for Super Trend – from AS-IS to TO-BE models using Bizagi Modeler. Implemented simulation, analysis and redesign to improve delivery efficiency and ROI through Zoho One system integration.

Hourly ride demand forecasting (weather + calendar) — NOVA IMS ML Bachelor's project

Customer segmentation for ABCDEats using advanced clustering (K-Means, SOM, Hierarchical) — NOVA IMS Data Mining 24/25

Customer segmentation for a food delivery platform | K-Means, Mean-Shift, Hierarchical Clustering | 4 segments | CRISP-DM | NOVA IMS 2024/25

Reinforcement learning for ICU sepsis treatment: tabular and deep RL with clinical failure-mode robustness and reward shaping.

Reproducible financial sentiment classification with NLP ensembles, transformers and knowledge distillation.

Genetic algorithm image reconstruction with perceptual colour metrics and statistically validated experiments.

NLP pipeline for financial tweet sentiment classification | RoBERTa, BERT, DistilBERT | Macro F1: 78.8% | NOVA IMS 2024/25

Portfolio fork of a collaborative Home Credit MLOps system with Kedro, MLflow, LightGBM, SHAP, drift monitoring, FastAPI, and Docker.

Deep-learning painter classification across 23 artists with transfer learning, duplicate auditing and 10-fold cross-validation.