Prajwal Sreeraman

Remote role focusing on data-driven decision making for enterprise analytics. Analyzed 1TB+ of customer, financial, and behavioral data using Python, MySQL, and Excel (Power Query, Pivot Tables) to improve data quality, feature engineering, and reporting accuracy. Built real-time Tableau dashboards by integrating live data streams from internal systems and automated ETL workflows with Excel macros, MySQL stored procedures, and AWS services, increasing data availability and pipeline throughput by ~60%. Led end-to-end analytics for a customer-recommendation initiative, translating model outputs into actionable insights that boosted CTR, loan conversions, and retention. Standardized core KPIs (engagement, CTR, CVR, retention, funnel metrics) across teams using Tableau dashboards and Excel KPI tools, reducing manual reporting time and enabling rapid stakeholder decision making.

Led end-to-end ML lifecycle for customer support using SV D/ALS hybrid POCs and high-scale data engineering (1 TB+ financial data). Implemented production-grade RAG agent to improve support accuracy and scalability; designed real-time data streaming architecture with Apache Kafka and AWS Kinesis; automated ETL pipelines (AWS S3, Glue, Redshift, EMR, SageMaker); built SQL-based validation layers to ensure data integrity and reduce discrepancies; defined core ML KPIs and deployed an in-house RAG application to enable contextual AI assistants.

Worked at the intersection of data analytics and pharmaceutical research to collect, clean, and analyze large-scale chemical and biological datasets. Built structured databases and automated ETL pipelines to support evidence-based decision making in drug discovery and R&D. Scraped and harmonized PubChem data using Selenium and BeautifulSoup, stored in MySQL, and deployed end-to-end data workflows in AWS. Implemented high-accuracy OCR for safety sheets using OpenCV, Tesseract, and TensorFlow, boosting text extraction accuracy by ~90%. Achieved 92% similarity accuracy in mapping 50K+ chemical compounds via fingerprinting (Morgan/ECFP, Atom Pairs) and multiple similarity metrics (Dice, Cosine, Tanimoto), enabling robust compound identification. Designed dashboards and reports to enable researchers and decision-makers.