Karthik RAmadugu

• Responsible for developing Machine Learning and Deep Learning driven Products and Services in domains such as Oil & Gas, Energy.
• Developed Reusable Assets/Programs and POCs for utilizing Parameter Efficient Finetuning Strategies such as Lora for finetuning and Inference of Large Language Models and Deep Neural Networks.

• Responsible for Developing CV & NLP Solutions and Reusable Assets Repository for Engineering Documents Mining, Image & Video Analytics Solutions to Automate Industrial Engineering Workflows.
• Responsible for Literature Survey of the Research Papers to understand the existing state of the Solutions and working upon them to modify and improvise solution tailored to the Domain.

• Developed End to End Reusable Python Modules for various Manufacturing and Engineering Services Workflows.
• Developed Computer Vision Pipeline for Automated Sewer Pipeline Defect Detection and Localization. The Solution utilizes Yolov5, Optical Flow, Midas-DPT and mask RCNN to perform Inspection and has resulted in 50% reduction in Time taken for Manual Inspection and Report Generation.

• Developed Deep Learning Models for Electronic Component Defect Classification that can detect the Expected Component Defect based on its Electrical, Mechanical and Manufacturing Process as Features and resulted in an average Per Class F1 Score of 0.8

• Developed Classification Model for Home Loan Default prediction.
• Developed Sentiment Classification Model to Analyze Tweets of Credit Card Products and draw insights about Sentiments over Time and Geography.
• Developed Home Loan Query Chatbot for Answering Queries regarding Home Loan Queries.