Data Science Intern at Wi-Tronix

(May 2024 - Aug 2024)

MLOps

Computer Vision

DL

Data Engineering

• Worked with stakeholders to optimize the ML model pre-processing pipeline, achieving up to 75% reduction in processing time and 70% decrease in RAM usage through distributed processing, significantly improving scalability
• Engineered ML Workbench to enhance the SDLC, reducing S3 storage costs by approx. $12000 and improving pre-processing efficiency for model training and testing while ensuring SLA compliance
• Developed custom deep learning model based on ShuffleNet and EfficientNet, enhancing transfer learning for detecting camera obstructions in train cockpit footage by 30%, and deployed it using Flask for production
• Developed end-to-end ML pipeline incorporating gradient boosting and random forests for initial feature selection, followed by custom deep learning models
• Implemented monitoring and analysis mechanism using Splunk, Grafana, and profiling tools to identify and resolve memory-bound issues, reducing consumption by over 3x
• Contributed to an advanced TTS system using Whisper, improving safety and reliability in high-noise environments
• Implemented robust data governance framework reducing data inconsistencies by 25% through comprehensive data lineage and validation techniques