Pushkaraj Palnitkar

Productivity Prediction ML Model

• Developed a multi-step machine learning algorithm to measure AWS Support-wide productivity across 4,000 engineers and 1 million resolved cases annually.
• Applied regression, supervised clustering, and SHAP analysis to identify product improvement opportunities, resulting in $4M projected annual savings.

Revenue Forecasting ML Model

• Developed a multi-step revenue forecasting algorithm using ensemble ML and deep learning techniques for Amazon's Just Walk Out.
• Improves forecasting accuracy by 40% resulting in improved revenue prediction with correcting >$2M revenue gap for reporting for Amazon leadership.

Enhanced operational processes to streamline workflows

• Designed and developed ETL pipelines and AWS QuickSight dashboards to identify and highlight critical business bottlenecks involving cross-team dependencies.
• Collaborated with cross-functional stakeholders (PMTs, TPMs, PMs) to design business reporting frameworks and resolve business metric ambiguities.
• Mentored junior team members and conducted hiring interviews for business intelligence teams, contributing to team growth and talent acquisition.

• Trained GAN (Generative Adversarial Networks) with CNN models using PyTorch on 3D MRA scans of over 100 patients.
• This yielded 6% improvement on previously tested U-Net architecture by researchers.

• Design, developed and deployed data analysis web application for material science research lab at University at Buffalo.

• Fine-tuned pre-trained BERT model for named entity recognition (NER) to extract key information from medical records.
• Delivered automated entity extraction from slip injury medical records, providing actionable insights to hospital management leadership for safety improvement initiatives.