Sarthak Srivastava Email & Phone Number

Brooklyn, New York, United States

•Solved Trend Analysis for retail clients by delivering GenAI driven model in 4 incremental steps:•Built Latent Semantic Indexing model, outperforming LDA-baseline by 30% improvement in accuracy in A/B Testing•Determined performance-related issues due to high latency of LSI model, deployed on AWS•Developed custom LLaMA 2 model that further reduced accuracy error rate by 30% and achieved 3 hours latency gain (4hrs to 1hr)•Packaged model in Docker and deployed on AWS EC2. Created REST API for easy integration in enterprise platforms•Optimized Machine Learning pipelines by refactoring Python code, implementing automated testing, and error handling resulting in a 60% reduction in model deployment time•Optimized data pipelines, integrating 100K+ records and reducing processing time by 35% using custom SQL/Python scripts•Leveraged pandas and Tableau to perform advanced data analysis and visualizations, uncovering insights that drove a 20% increase in strategic decision-making accuracy

Brooklyn, New York, United States

Pune, Maharashtra, India

•Achieved 0.82 mean ROC-AUC for personal attribute classification by building a based ResNet-50 architectures using Python & TensorFlow 2•Corrected incorrect video-frames by adding a orientation correction model, achieving below 4° mean angular error•Built a 2-stage classification pipeline with frame correction downstream model, that improved upstream classification model accuracy by 20%

Bengaluru, Karnataka, India

•Elevated data integration processes leveraging SQL Server and Python for multi-source air quality data from sensors, weather stations, and trafficmonitors. The standardized and automated ETL resulted in a 40% improvement in data preparation time•Applied time-series analysis to build machine learning models identifying trends and anomalies in air quality indicators, showcasing expertise intime-series modeling•Pioneered the implementation of MLOps practices, including CI/CD pipelines for model deployment, resulting in a 60% reduction in model releasetime. This facilitated quicker adaptation to changing environmental conditions•Collaborated with cross-functional Agile teams across air quality monitoring, operations, and reporting. Streamlined processes delivered a 10%operations cost reduction and a 22% faster regulatory reporting process

Nagpur, Maharashtra, India

Deep Learning Researcher | Emotion Recognition Solutions- Engineered a Deep Convolutional Neural Network, achieving an impressive 92% accuracy in classifying facial expressions into seven emotional categories. This technological advancement significantly boosted our market edge.- Conducted meticulous hyperparameter tuning, resulting in a remarkable 22% improvement in classification accuracy over baseline models.- Leveraged cross-functional collaboration to drive emotion recognition insights in healthcare and HCI, contributing to a strengthened market position and spurring strategic product innovations.

Master Of Science - Ms, Information System

Coursework : Application Engineering Development Data Management and Database Design Data Science Engineering Methods and Tools Web.

Bachelor Of Engineering - Be, Electronics Design Technology