Saili Myana
Data Scientist & AI Research Engineer
Transforming complex data into actionable insights with cutting-edge AI models. Enhanced forecasting accuracy by 200bps at Walmart | MS Data Science from Columbia University
About Me
I'm a data scientist passionate about leveraging AI and machine learning to solve complex business problems. With an MS in Data Science from Columbia University (4.12 GPA) and BTech from IIT Madras, I specialize in demand forecasting, predictive modeling, and deep learning applications. Currently enhancing forecasting systems at Walmart using advanced AI models.
Education
Core Expertise
Key Achievements
Professional Journey
Data Scientist
Enhanced forecasting accuracy by 200bps using AI models like DeepAR, Prophet, ARIMA, LSTM. Developed 8 data pipelines using Airflow and Kubeflow on GCP for weekly updates.
Data Science Intern
Identified key item attributes through correlation analysis of 10 categories. Implemented like-item generation service providing 5.5% improvement in forecasts.
AI Research Intern
Customized AC-TPC deep learning model for predictive user segmentation. Applied for US patent in Predictive Profiling.
AI Research Intern
Proposed Double U-net architecture for Medical Image Segmentation with 0.71 dice-similarity coefficient using half the parameters. Published research paper at ICVISP 2020.
Skills Evolution: Research to Production
Key Projects
J.P. Morgan: Editing Large Language Models (Capstone)
Investigated methodologies for editing facts in GPT 2-XL model with 1.5b parameters. Identified knowledge layers using Causal Tracing, demonstrating facts about popular and rare entities stored in different layers.
Walmart Demand Forecasting System
Production AI system enhancing forecasting accuracy by 200bps using DeepAR, Prophet, ARIMA models. Manages 8 automated pipelines for real-time model training.
Adobe Predictive Customer Segmentation
Customized AC-TPC deep learning model with attention layers in LSTMs. Applied for US patent in Predictive Profiling for advanced customer behavior analysis.
Medical Image Segmentation Research
Double U-net architecture with Modified Swish activation achieving 0.71 dice-similarity coefficient with 50% fewer parameters. Published at ICVISP 2020.
Let's Build Something Amazing Together
Ready to tackle your next data science challenge? I specialize in AI-driven forecasting, deep learning research, and production ML systems. From prototype to production, I deliver measurable results.