Projects

Technologies: Python Transformers

A systematic exploration of transformer architecture fundamentals, approaching neural attention mechanisms through the lens of physics intuition and structured analysis. This project documents my methodology for rapidly understanding complex ML architectures and identifies physics-inspired research questions for future investigation.

Key Insights:

• Physics-informed understanding of attention mechanisms
• Systematic learning framework for complex architectures
• Hands-on exploration with conceptual testing
• Physics-inspired research questions for future investigation

Read More

Technologies: Python XGBoost GCP/Vertex.ai Recommendation Systems

Production ML system for personalized load recommendations at CloudTrucks, improving driver load booking rates through intelligent matching. Built end-to-end pipeline from feature engineering to real-time inference, serving hundreds of drivers daily.

Key Achievements:

• 12% increase in driver load booking rate through personalization
• XGBoost algorithm selection and optimization for recommendation system
• Data warehouse pipeline design for ML feature engineering
• A/B testing framework and continuous model monitoring

Read More | CloudTrucks Engineering Blog

Technologies: Python PyMC3 Bayesian Inference Maritime Analytics

Automated sea condition detection for cargo ships using time series sensor data and Bayesian logistic regression. Developed during Insight Data Science Fellowship to address small data challenges in maritime consulting applications.

Key Achievements:

• Bayesian approach handling uncertainty with small data
• Physics-inspired feature engineering for sensor data analysis
• Successful model validation using posterior predictions
• Automated reporting foundation enabling future fuel optimization

Read More