Dec 2024 – Present · 1 yr 6 mos

• Optical Character Recognition (OCR) Project:
• Currently working on an OCR project aimed at revolutionizing text extraction and processing from diverse sources, with a specialized focus on interpreting complex handwriting. Key aspects of the project include:
• Automated Text Extraction: Designing and fine-tuning algorithms to accurately extract text from scanned documents, handwritten notes, and images, including challenging cases like doctor handwriting.
• Handwriting Recognition: Developing advanced models tailored to decipher and interpret medical prescriptions and notes written by doctors, addressing the nuances of irregular and stylized handwriting.
• Preprocessing Techniques: Implementing robust preprocessing methods like noise reduction, skew correction, and image binarization for improved accuracy and reliability.
• Machine Learning Integration: Leveraging deep learning techniques, including convolutional neural networks (CNNs) and recurrent neural networks (RNNs), to enhance recognition capabilities, even for poorly written or low-quality input.
• Multi-Language and Contextual Understanding: Expanding OCR capabilities to handle multiple languages and incorporate contextual understanding, particularly for medical terms and abbreviations.

Dec 2024 – Present · 1 yr 6 mos

• I worked on language modeling projects, focusing on developing and fine-tuning NLP models for various text-processing tasks. My responsibilities included:
• Data preprocessing: Cleaning, tokenizing, and formatting large text datasets for training and evaluation.
• Model development: Experimenting with transformer-based architectures (e.g., GPT, BERT) and traditional language models to improve accuracy and efficiency.
• Hyperparameter tuning: Optimizing model performance through experimentation with learning rates, batch sizes, and architectures.
• Evaluation & analysis: Measuring model accuracy, perplexity, and other NLP metrics to ensure robustness.
• Collaboration: Working closely with the AI team to brainstorm ideas, troubleshoot issues, and integrate models into pipelines.