Sowmya Kunam

Results-driven Data Analyst with a strong analytical mindset and expertise in transforming complex datasets into actionable insights. Proven ability to drive informed decision-making through data interpretation, visualization, and effective communication.

Fake News Detection

• Curated and preprocessed a diverse dataset of news articles, applying natural language processing (NLP) techniques.
• Engineered textual features, including TF-IDF and word embeddings, to capture semantic nuances and improve model accuracy.
• Developed and fine-tuned machine learning models, such as Random Forest and Multinomial Naive Bayes, for classification.
• Implemented cross-validation strategies to assess model generalization and robustness.
• Utilized ensemble techniques, combining multiple models to enhance overall prediction performance.
• Conducted thorough model evaluation, including precision, recall, accuracy and F1-score metrics, ensuring reliable performance.
• Applied the best model on test dataset to find the Fake News.

Food  Image Detection

• Conducted data preprocessing, cleaning, and augmentation to optimize the dataset for deep learning model training.
• Implemented a convolutional neural network (CNN) architecture using TensorFlow and Keras for accurate food image recognition.
• Utilized transfer learning techniques with pre-trained models (e.g., VGG16, ResNet) to leverage existing knowledge for improved performance.
• Fine-tuned hyperparameters, conducted cross-validation, and implemented regularization techniques to enhance model generalization.
• Collaborated with UX/UI designers to ensure a seamless and intuitive user experience.
• Conducted comprehensive testing and evaluation to validate model accuracy and identify areas for improvement.
• Implemented continuous monitoring and model updates to adapt to evolving data patterns and improve overall performance.