Matheus Cardoso

Led development of a financial data solution unifying receivables (including future quotas), advances, and cash-flow views to support the Financial Risk team’s decision-making. Developed Expected Credit Loss (ECL) modeling for Mexico, including market simulators and expected loss/risk analyses for regulatory compliance.

Built an advanced chatbot integrating Gemini and Claude LLMs, using NER-based data anonymization for privacy-compliant handling of user data. Implemented federated learning with LLMs and knowledge graphs to preserve confidentiality while improving personalization.

Built a vector-based chatbot (LLM + FAISS + metadata) as a Data Discovery Bot that generated R$2M in revenue for the company. Deployed a real-time NPS classifier integrating usage and billing data to support retention strategies.

Developed GCP-to-AWS pipelines for revenue and chargeback reconciliation. Created an Airflow-based data-quality validation system to improve robustness in high-volume pipelines.

Applied Transformer/BERT models to streamline inspection report analysis. Built a Spark-based model to detect transformer failures in ~1B IoT records per hour, reducing downtime.

Developed asset-based analysis models for data failure prediction and clustering using IoT data. Built data processing workflows capable of handling over 1B lines of IoT data for downstream analytics.

Master Data Science and Analytics with a thesis on applying LLMs to help a startup business plan using the Socratic method.

Post graduation in Machine Learning and Artificial Intelligence.

Electrical Engineer with an electronics emphasis.