Vraj Shah

Designed and implemented end-to-end pipelines for Single Line Diagram (SLD) analysis, automating electrical component detection and reducing manual effort by 30%. Deployed scalable models in AWS EC2 using MLOps practices, enabling automated production pipelines. Developed real-time object detection using YOLO with 95% accuracy, enhancing video processing capabilities. Applied time-series forecasting models to improve decision-making in operational processes of energy distribution. Collaborated on building a chatbot using LLAMA-2 framework, reducing customer query resolution time by 40%. Built Retriever Augmented Generation (RAG) systems to integrate external data for enhanced question-answering capabilities. Enhanced pattern identification accuracy by 25% through image segmentation-based curve detection algorithms. Utilized GANs for synthetic SLD image generation, improving data diversity and training outcomes. Implemented Graph Neural Networks (GNN) to predict connections in SLD creation, streamlining diagram design.

Promoted to Algorithm Developer after excelling as an intern, demonstrating strong technical and problem-solving skills. Designed algorithms for Advanced Driver Assistance Systems (ADAS) to classify obstacles using machine learning models, achieving 98% accuracy. Developed LSTM-based models to analyze radar data for obstacle detection and classification. Classified various objects, including pedestrians, vehicles, and infrastructure, enhancing autonomous vehicle perception capabilities. Innovated specialized observers to classify obstructions into under-ridable, over-ridable, and obstacles for improved safety in driving scenarios.