Sai Pravallika Pasumarthi

Health AI – Intelligent Healthcare Assistant

• Built an AI-driven healthcare assistant using MedAlpaca-13B, Python, and Streamlit, supporting medical Q&A, symptom-based diagnosis, and treatment recommendations.
• Integrated Hugging Face transformers pipeline to process 50+ symptom types and generate context-aware medical responses with improved latency.
• Developed an interactive health analytics dashboard using Plotly and Pandas, visualizing 10+ vital parameters and generating AI-based insights.
• Architected a modular 4-component system (Chat, Diagnosis, Treatment, Analytics) with secure .env token management, reducing code complexity by 30%.

Fake Image Detection using Deep Learning

• Developed a binary image classification model using EfficientNet-B0 and PyTorch, detecting AI-generated vs real faces.
• Applied transfer learning with advanced preprocessing (resizing, normalization) and data augmentation (rotation, zoom, flip), improving robustness on a 20k+ image Kaggle dataset.
• Achieved 94% accuracy, 92% precision, 91% recall, and 92% F1-score on the test set using optimized training and regularization.

Built and optimized ML models achieving up to 90% accuracy, preprocessed large datasets (5K+ records), and integrated LLM APIs to reduce manual query time by 50%.

• Conducted research for various projects, leading to well-informed decisions and successful outcomes.