Siddhi Gunaji

• Built Python-based data pipelines on AWS to ingest, transform & validate fragmented vehicle safety datasets, resulting in reliable, analytics & ML-ready data products. • Developed automated dashboards in AWS QuickSight, consolidating outputs from multiple AI workflows into actionable insights for engineering decision-making . • Contributed to computer vision–based vehicle safety use cases, like driver attention monitoring, supporting data preparation & evaluation for in-cabin perception systems.

• Designed & operated production-grade time-series forecasting pipelines using the Nixtla ecosystem to support scalable demand & KPI forecasting. • Implemented feature engineering, automated validation, & performance monitoring workflows, improving forecast consistency, reliability, & reproducibility across products and regions. • Collaborated with product & business stakeholders to translate planning requirements into reusable forecasting & analytics workflows, enabling data-driven decision-making.

• Analyzed large-scale time-series sensor & PLC data to detect early failure patterns, applying statistical forecasting models, LSTM, XGBoost to support predictive maintenance & operational monitoring. • Built end-to-end data preprocessing & feature engineering pipelines, & performed model training, validation, & performance comparison to deliver robust & interpretable analytical results. • Documented assumptions, thresholds, & evaluation results, ensuring traceability, and repeatability.

• Extracted meaningful insights from over 4.6 million records of PLC signal data using advanced big data methods like NumPy arrays, Dask, and Polars, enhancing data understanding and predictive maintenance capabilities. • Performed exploratory data analysis (EDA) and applied multiple machine learning models on cutting machine tool data in the engine stator production station, achieving approximately 72% accuracy in predicting blade replacement timings and improving production efficiency.

• Engineered and optimized tax fraud detection systems using Apache Kafka, Spark, and Airflow for real-time data ingestion and ETL, and deployed anomaly detection models with TensorFlow and AWS SageMaker, enhancing fraud detection and system efficiency. • Developed and maintained SQL queries and stored procedures using Teradata at the Central Board of Direct Taxes, leveraging Big Data expertise for effective data management and analysis.