- Fine-tuned PaddleOCR on custom datasets for efficient text extraction from images, enabling accurate data collection. - Fine-tuned and evaluated DistilBERT for brand name and personal information recognition in text. - Trained an XGBoost model from scratch for the categorization of product-related PDFs. - Built AI agents using LangChain to automate the collection and storage of product metadata for seamless database integration. - Developed a RAG-based chatbot using… Show more - Fine-tuned PaddleOCR on custom datasets for efficient text extraction from images, enabling accurate data collection. - Fine-tuned and evaluated DistilBERT for brand name and personal information recognition in text. - Trained an XGBoost model from scratch for the categorization of product-related PDFs. - Built AI agents using LangChain to automate the collection and storage of product metadata for seamless database integration. - Developed a RAG-based chatbot using LangChain, OpenAI, and FastAPI to extract insights from product-related PDFs. - Created a semantic search module for fast, context-aware retrieval of relevant documents. - Designed an asynchronous pipeline using Celery, RabbitMQ, and Redis for efficient task execution, caching, and coordination of large-scale data retrieval. - Integrated Elasticsearch to enable fast and accurate search functionality for complex queries. - Developed a user-friendly API with FastAPI, providing seamless access to aggregated data in Excel or CSV formats. Show less

- Optimized Emotion Detection Model: Achieved 2x faster inference by integrating Gunicorn with Docker, enabling real-time processing of over 1,000 tasks per second. - Enhanced Face Detection Performance: Reduced inference time by 30% through model quantization, while maintaining a high mAP of 97.8%, significantly improving tracking efficiency.

- Designed and implemented advanced inspection systems to enhance defect detection and quality control. - Used a Siamese network with PyTorch to detect missing parts in engine inspections with high precision. - Trained a YOLOv4 model for identifying defects in washing machine drums, improving detection speed and accuracy. - Optimized an OCR solution for character recognition on LPG cylinders, achieving 98.2% precision. - Developed a cap inspection system with 99% precision… Show more - Designed and implemented advanced inspection systems to enhance defect detection and quality control. - Used a Siamese network with PyTorch to detect missing parts in engine inspections with high precision. - Trained a YOLOv4 model for identifying defects in washing machine drums, improving detection speed and accuracy. - Optimized an OCR solution for character recognition on LPG cylinders, achieving 98.2% precision. - Developed a cap inspection system with 99% precision, enhancing production efficiency and ensuring safety. Show less