Hi, I’m Bruno Chicelli
AI/ML Engineer

Results
Creating the best AI to estimate remodeling costs.

Results
Worked on the core title-matching algorithm to accomplish:
• 750× performance improvement (from 3s/title to 4ms/title)
• Significant precision gain in matching movies, shows, and episodes
• Scaled to 3M+ titles across 60+ streaming services in 8 languages

Results
• Improved ad recommendation models using Two Towers architecture, matching/beating expert performance
• Scaled behavior graph models and pipelines for 100M+ users using Vertex AI, DataProc, DataFlow, and BigQuery
• Won internal hackathon with user embeddings + generative AI for segment-optimized ad creatives

Results
• Built automated midroll ad placement pipeline for thousands of Globo videos using event-based cloud tools and deep learning
• Developed contextual ad engine with text-to-IAB taxonomy matching and topic clustering, significantly improving CTR
• Founded Globo Paper Club, a monthly internal journal club
• Led projects in digital inventory optimization, pricing, and sales forecasting

Results
• Built a fast modeling tool with LGBM, XGBoost, and CatBoost to reduce Data Scientist overhead
• Developed a credit card acceptance and activation propensity model
• Improved modeling process standards
• Led MLOps and workflow automation innovations to boost impact, productivity, and enablement of the Data Science team

Results
• Saved ~1.3M BRL/year through automation and data science
• Built sales forecasting for 100+ points of sale across Brazil
• Developed customer churn and factory demand prediction models
• Applied statistical modeling and ML in Python and R for forecasting and insights
• Contributed to analytics projects across multiple business areas