Satyam Kesharwani

Co-developed a Residual-Physics Neural Network (RPNN) embedding Coulomb Counting equations into the loss function for battery State-of-Charge estimation, achieving R2: 99.70% Deployed an INT8-quantized TFLite Micro model on ESP32 microcontroller, embedding the model binary into firmware via C header generation, reducing model size by 4x while maintaining R2: 99.70% Invented a "Virtual Cranking" algorithm to simulate engine-start loads and predict vehicle start failure, improving system reliability for real-world IoT deployments

Software Engineering Intern (ML Systems & Edge Deployment) - National Institute of Technology Kurukshetra - Kurukshetra, India

(2025-06)

• Co-developed a Residual-Physics Neural Network (RPNN) embedding Coulomb Counting equations into the loss function for battery State-of-Charge estimation, achieving R2: 99.70%
• Deployed an INT8-quantized TFLite Micro model on ESP32 microcontroller, embedding the model binary into firmware via C header generation, reducing model size by 4x while maintaining R2: 99.70%
• Invented a "Virtual Cranking" algorithm to simulate engine-start loads and predict vehicle start failure, improving system reliability for real-world IoT deployments
• Built an end-to-end IoT data pipeline with real-time cloud dashboard for battery health monitoring, handling continuous sensor telemetry streams; presented at IEEE NE-IECCE 2026

Software Engineering Intern (ML Infrastructure) - National Institute of Technology Kurukshetra - Kurukshetra, India

(2024-06 - 2024-12)

• Built RansomDroid, an end-to-end Android malware detection system integrating Vision Transformer (ViT) and CNN; identified ViT as a superior approach after initial CNN baseline underperformed on obfuscated samples, achieving a 4% accuracy gain
• Engineered a scalable data ingestion pipeline in Python to parse 4,280 APKs, automating JSON-to-CSV extraction with optimized I/O and batch processing
• Designed a high-throughput inference pipeline with batch-optimized processing, achieving 99.78% detection accuracy across large-scale Android application datasets
• Published results as lead author at ICDAM 2025 (Springer Lecture Notes in Networks and Systems)

Bachelor of Technology - Computer Engineering - National Institute of Technology, Kurukshetra (2023 - 2027)