I am an engineer (currently pursuing my master’s in Interaction Technology) with a strong focus on natural language processing, conversational systems and applied AI. I have hands-on experience developing deep learning models and working with multimodal data. I enjoy learning new things, adapt quickly to challenges and work well with people. I am curious about emerging trends in technology and I am always open to learning more, especially in the fast-evolving fields of NLP and AI. With a disciplined yet creative approach, I aim to build intelligent systems that are both technically sound and people-focused.
- Completed a one-month training program from August to September.
- Worked on two machine learning projects:
- MNIST Digit Classification (handled independently).
- Music Genre Classification (team project).
- Applied machine learning techniques to improve classification accuracy.
- Developed an SMTP client application using Python.
- Created and delivered eight versions of the application.
- Improved email handling efficiency through iterative development.
Predicting the Type and Target of Offensive Posts in Social Media
• Built a multi-level classifier to detect and label offensive tweets using the OLID dataset.
• Compared SVM, BiLSTM, RNN, and BERT meta-learning models for performance on text classification tasks.
• Achieved up to 77% accuracy with contextual feature extraction and advanced preprocessing.
Residual Learning of Deep CNN for Image Denoising
• Applied DnCNN architecture to denoise images affected by Gaussian, salt & pepper, and speckle noise.
• Used residual learning and batch normalization to enhance training convergence and image quality.
• Evaluated denoising performance using PSNR and SSIM on the Kodak dataset.
Text Summarization Model for Tamil Language using LLM
• Built a Tamil text summarization system by fine-tuning Llama2 (7B) using Low-Rank Adaptation (LoRA) techniques.
• Achieved ROUGE-1 (~56%) and ROUGE-2 (~30%) scores, outperforming baseline models like GPT-2 and BLOOMZ.
• Addressed challenges in low-resource language modeling through careful dataset curation and evaluation using ROUGE and BLEU metrics.
At the Integrated Intervention Center in Bahrain, I volunteered to support teachers in managing classroom activities for children with special needs, and assisted in coordinating school events. This experience gave me meaningful insight into inclusive education and helped me better understand how to support children with unique learning and developmental needs