Designed and maintained 20+ scheduled production data pipelines and analytics workflows using AWS, Databricks, Python, SQL, and Delta Lake. Developed full-load ingestion, transformation, validation, and delivery workflows connecting operational systems with analytics-ready datasets. Modeled financial and operational data across Bronze, Silver, and Gold layers, applying schema standardization, deduplication, joins, and business-rule mapping. Automated data extraction, Excel/CSV conversion, validation, and S3 delivery using Python, Selenium/RPA, and boto3. Built configuration-driven workflows supporting multiple client environments, processing periods, source mappings, and S3 destinations without changes to core logic. Implemented secure data-lake access using Amazon S3, IAM roles, bucket policies, and role-based access controls. Maintained AWS Lambda functions for scheduled full-load processing, aligning execution frequency with data availability and operational cost. Collaborated with operational teams to translate heterogeneous source-system requirements into reliable analytical models and reporting datasets.

Curated and processed a bilingual dataset of 30,000+ English and Portuguese image-caption pairs aligned with FrameNet semantic structures. Fine-tuned YOLOv8 models for frame-conditioned object-detection experiments and adapted BLIP models for frame-aware image captioning. Developed reproducible experiments combining computational linguistics, computer vision, and multimodal machine learning. Built and published interactive machine-learning demonstrations using Hugging Face and Gradio.

Automated social-media data collection and reporting workflows using Python, Selenium, and RPA, eliminating an estimated 500+ hours of repetitive work annually. Developed Python and Plotly dashboards to monitor campaign performance, sentiment trends, engagement, and audience behavior. Processed and analyzed social-media datasets to identify patterns and support campaign and communication decisions.