Yashwani Patidar

Challenge: Fragmented garage booking workflows and lack of secure digital payment tracking.

Impact: Built a robust REST API ecosystem with OAuth2/JWT security. Integrated Razorpay for seamless transactions and AWS S3 for secure document storage. Reduced booking friction by 40%.

Challenge: Critical electrical panel failures leading to massive unplanned industrial downtime.

Impact: Deployed a YOLOv8-based thermal fault detection pipeline. Features real-time OCR extraction of temperature data from thermal images. Currently hosted on HuggingFace Spaces.

Challenge: Standard LLMs suffer from hallucinations in specific travel domain knowledge.

Impact: Designed a modular RAG system combining Vector Search (FAISS) for semantic retrieval and Graph Database (Neo4j) for relational context. Optimized for high-precision responses.

Challenge: Identifying AI-generated facial manipulations in low-resolution media.

Impact: Developed a binary classifier using ResNet50. Implemented custom data augmentation and fine-tuning to achieve high sensitivity against GAN-generated artifacts.

Challenge: Sub-optimal CNN hyperparameters affecting classification.

Impact: Bio-inspired Meta-Heuristic algorithms (Genetic Algorithm & Wolf Optimization) for architecture and LR tuning. Achieved 94.79% accuracy on 47-class EMNIST (130K images) with automated architecture & learning rate selection.

Challenge: Manual attendance tracking inefficiency.

Impact: Deployed face recognition-based automation with real-time face recognition using live camera. Captures 40 facial samples per person for accuracy, stores data in SQLite database with referential integrity, and includes automated report generation system.