Pirathiish Rajasegaran

Recent graduate with excellent research, technical, and problem-solving skills. Detail-oriented and able to learn new concepts quickly.

Artificial Intelligence Analyst- IBM,

• Specialized in NLP, Computer Vision, and Chatbots, focusing on their application in data analysis and modeling. Proficient in leveraging IBM tools like Watson, Discovery, and Knowledge Studio for developing AI applications.

Google Cloud Computing Foundations

• Gained a solid understanding of cloud computing, big data, and machine learning concepts with a focus on Google Cloud Platform. Completed hands-on training via Qwiklabs to apply and practice key skills in a real-world environment.

Enhanced Medical Image Classification under Resource Constrained Scenarios

• Developed a feature selection algorithm by combining Reinforcement Learning (Double Q-Learning) with Genetic Algorithms to enhance model performance, particularly in low-data scenarios
• Applied this approach to tumor classification using CT and MRI images, optimizing feature selection for both ML and DL models.
• The method demonstrated improved accuracy by selecting features that maximize model performance in image recognition tasks.

Web Application for Image Classification using Transfer Learning

• Experimented with multiple transfer learning models for image classification, tracking performance using Weights & Biases and TensorBoard.
• Analyzed model metrics against various parameters to select the optimal model, which was then deployed as a web application using Gradio for real-time image classification.
• various speech emotion recognition techniques, treating the problem as an image recognition task. Leveraged transfer learning algorithms, tested performance using TensorFlow and Weights & Biases, and performed hyperparameter optimization to enhance accuracy. Analyzed and combined speech features to improve model performance.

Speech Emotion Recognition with Transfer Learning

• Investigated and applied various speech emotion recognition techniques, treating the problem as an image recognition task.
• Leveraged transfer learning algorithms, tested performance using TensorFlow and Weights & Biases, and performed hyperparameter optimization to enhance accuracy. Analyzed and combined speech features to improve model performance.