SWETHA A

Aspiring Artificial Intelligence and Machine Learning enthusiast with a strong foundation in Python programming and hands-on experience building neural networks and CNNs from scratch. Actively expanding knowledge in deep learning, computer vision and model optimization through personal projects. Certified in Pega System Architect (CSA & CSSA) and familiar with tools like Flask and SQL. Eager to grow as an AI/ML developer and contribute to real-world, data-driven solutions as a fresher in the field.

Python Basics for Data Science - IBM edX

Project HorticBot : Integrated Pest Management & Crop Health Evaluation

• Designed and developed an autonomous agricultural rover using Raspberry Pi 4 with a Python-based control system integrating OpenCV for real-time image processing.
• Implemented Convolutional Neural Networks (CNNs) based on VGG16 architecture for Pest Detection (96.44% accuracy) and Leaf Disease Identification (80.67% accuracy).
• Developed Weed Detection module using YOLO (You Only Look Once) object detection model with Non-Maximum Suppression (NMS) for precision filtering (90% precision).
• Built custom NDRE Analysis module by engineering a multispectral sensor (Raspberry Pi camera + IR filter) to calculate Normalized Difference Red Edge (NDRE) values for chlorophyll content analysis.
• Technologies used: Python, OpenCV, TensorFlow/Keras, YOLO, Raspberry Pi OS, Flask (future scope for dashboard integration).

Facial Emotion Detection using CNN from Scratch

• Developed a multi-class facial emotion classifier using a manually implemented Convolutional Neural Network without external libraries.
• Processed the FER2013 dataset (48x48 grayscale images) to classify 7 emotions: Angry, Disgust, Fear, Happy, Sad, Surprise, Neutral.
• Implemented core CNN components manually: convolution layers, max pooling, flattening, fully connected layers, softmax and cross-entropy loss.
• Applied mini-batch stochastic gradient descent and backpropagation with separate validation evaluation.
• Achieved 83% model accuracy on the test set.

Parkinson’s Disease Detection using Neural Networks

• Built a binary classification neural network to detect Parkinson’s disease from biomedical voice data (UCI dataset).
• Applied feature scaling, trained the model using Keras, and evaluated performance using accuracy/loss plots.
• Implemented model persistence and a prediction interface to classify new patient data.
• Achieved up to 95% test accuracy and analyzed overfitting behavior through validation trends.
• Tech Stack: Python, TensorFlow, Scikit-learn, Pandas, NumPy.

Project Medicure :

• Developed a Pega-based platform that enables customers to book medical tests at nearby lab centers or at their residence.
• Designed an intuitive interface for seamless scheduling and automated notifications.
• Leveraged Pega’s workflow automation to streamline appointment management and enhance user experience.

• Published paper titled "Enhancing Crop Health Monitoring and Disease Identification in Agriculture" – 2023 Intelligent Computing and Control for Engineering and Business Systems (ICCEBS), IEEE
• Published paper titled "Advanced NDRE Techniques for Precision Crop Monitoring and Analysis" - 10th IEEE International Students’ Conference on Electrical, Electronics and Computer Science
• Project HorticBot – Patented and Published