PRAJWAL SINGH

Aspiring Data Scientist with a strong foundation in Python, machine learning, and data visualization, aiming to apply statistical analysis, predictive modeling, and natural language processing (NLP) to extract actionable insights and solve real-world business problems.

• Neural Networks and Deep Learning (DeepLearning.AI): Gained foundational knowledge of deep learning, including neural network architecture, forward and backward propagation, and applications in image and text classification.
• Improving Deep Neural Networks: Hyperparameter Tuning, Regularization and Optimization (DeepLearning.AI): Learned techniques to enhance neural network performance, including hyperparameter tuning, applying regularization methods such as dropout and L2, and using optimization algorithms like Adam and RMSprop to improve training efficiency and reduce overfitting.

Intelligent Chatbot Development: Built an NLP-based chatbot using cosine similarity to match user queries with relevant responses, ensuring contextual accuracy through a custom similarity threshold. Integrated Groq API for high-speed language processing and managed real-time data fetching and response generation via REST APIs. Applied text vectorization and semantic matching to improve chatbot responsiveness.

Defect Detection in Mobile Phone Images using Deep Learning (TensorFlow): Built an image classification system using TensorFlow and Keras to detect physical defects (e.g., screen cracks, dents) in second-hand mobile phone images for resale platforms. Designed a custom Convolutional Neural Network (CNN) with Conv2D, MaxPooling, and Dropout layers for robust feature extraction. Leveraged data augmentation and early stopping to improve model performance on varied backgrounds and lighting. Achieved over 92% validation accuracy, enabling scalable automation of quality checks for platforms like Cashify or OLX.

Libranet, Digital Library System: Developed a web-based inventory management system using the Django framework to manage books, users, and borrow/return activities. Optimized data handling for better resource allocation and provided user access controls, improving efficiency in library operations.

Autonomous Support Agent using LLMs and Reinforcement Learning: Built an AI support assistant using LLMs, Reinforcement Learning (RLHF), and agent-based architecture for goal-driven conversations. Enabled multi-step reasoning and context retention with LangChain, custom reward functions, and deployed using OpenAI API and Streamlit.