Taro Shibusawa

https://www.verif.com/ For this Company, led AI/ML initiatives for enterprise-scale predictive analytics and intelligent automation solutions - Predictive Analytics Platform: Architected a company-wide platform processing 2TB+ of daily data from 5M+ users. Built distributed ETL pipelines with Spark & Airflow, and deployed XGBoost models for churn prediction, achieving an 18% YoY reduction in customer attrition and $1.5M in annual revenue retention. - Full-Stack Data Visualization: Led end-to-end development of an enterprise dashboard using React, TypeScript, and D3.js. Engineered a FastAPI backend, reducing average response time from 3.2s to 0.4s. Integrated Kafka for real-time streaming, eliminating 20+ hours/week of manual reporting. - NLP & Automation: Pioneered a transformer-based NLP system to automate categorization of 500K+ annual support tickets. Fine-tuned DistilBERT (92% accuracy), deployed via TorchServe on AWS ECS, reducing resolution time by 35% and improving customer satisfaction by 22%. - Leadership: Mentored 3 junior engineers, established team coding standards, and improved code quality metrics by 45%.

For this Company, drove AI-powered innovation in agricultural technology through computer vision and predictive modeling. Computer vision system: designed and deployed a CNN-based defect detection system using YOLOv8, achieving 99.2% accuracy. The system saved over 1,200 monthly labor hours and reduced defect-related losses by 85%. Supply chain forecasting: developed time-series forecasting models (LSTM, ARIMA, Prophet) using PySpark, improving forecast accuracy by 34% and reducing supply chain waste by 20%. AI-powered decision support: created a GPT-3-powered chatbot with RAG architecture for over 3,000 farmers, providing expert advice that reduced crop losses by 15%. Mobile & IoT: Built a Flutter app with on-device TensorFlow Lite inference for over 5,000 users. Architected an IoT data infrastructure connecting more than 500 field sensors.

For this Company, applied statistical analysis and ML to pharmaceutical research and clinical trials. Clinical trial analysis: performed advanced statistical analysis on phase III trial data for over 10,000 patients. Used Bayesian methods for adaptive trial design, contributing to a four-month acceleration in FDA approval timeline. Predictive modeling: developed LSTM/Transformer models to forecast patient enrollment, achieving 87% accuracy and reducing operational costs by 15% ($2.8M savings). Data infrastructure: architected a scalable data model in Snowflake and built automated ETL pipelines with Airflow, reducing query execution time by 75%.