SASIDHAR CHERIVI

Project A: SoC Prediction and Battery Test Bench Development for TEMI Robot Battery System

• Developed SoC prediction models for the TEMI robot battery system, improving prediction accuracy by 15% and reducing forecasting error by 18%.
• Built an ESP32-based battery monitoring system using INA260 and MCP9808 sensors to measure current, voltage, power, and temperature.
• Integrated battery data with InfluxDB for real-time storage, querying, and time-series analysis.
• Developed Python scripts for battery data collection, preprocessing, and experimental evaluation.
• Supported battery test bench development and validation, including setup, monitoring, and system documentation
• Languages/Technologies: Python, C++, ESP32, InfluxDB, INA260, MCP9808, Time-Series Forecasting, DOE, Wi-Fi Logging

Project B: Deployment of Anomaly Detection Systems Using Machine Learning.

• Engineered and deployed production-ready ML models for anomaly detection systems using TensorFlow and PyTorch, boosting prediction accuracy by 18%.
• Orchestrated end-to-end ML pipelines with model monitoring to address overfitting and imbalanced datasets, increasing the F1 score from 0.72 to 0.85.
• Streamlined data preprocessing workflows and ensured data versioning with MLFlow, reducing pipeline processing time by 30%.

• Developed and deployed production-ready ML APIs with Flask, supporting 100+ concurrent users with high reliability.
• Automated model retraining workflows for document classification systems, reducing manual effort by 30% and improving operational efficiency.
• Enhanced model lifecycle management using MLflow, enabling traceability across 12+ projects and shortening update cycles from 48 hours to 3 hours.
• Built scalable deployment pipelines with Docker, Kubernetes, and CI/CD, improving scalability by 25% and cutting deployment time by 40%.

Languages/Technologies: Python, Flask, MLflow, Docker, Kubernetes, CI/CD, Machine Learning, Document Classification