Developed an intelligent Virtual Sales Agent that seamlessly simulates realistic customer interactions. Leveraging LangChain, LangGraph, and SQLite database, this agent provides comprehensive product information, processes orders, checks order status, and explains company policies. The system reduces customer service response time by 40% while maintaining high customer satisfaction levels through personalized responses and 24/7 availability.

Constructed a deep neural network model implementing a CRNN with CTC loss for Vietnamese speech recognition. Systematically collected and prepared the VIVOS dataset, applying speech processing techniques to optimize input data. Achieved an average Character Error Rate (CER) of 18.75% and a Word Error Rate (WER) of 46.64%. Additionally, fine-tuned the OpenAI Whisper-small model on the VIVOS dataset, achieving an improved WER of 16.38%.

This project demonstrates a real-time data pipeline that streams data from Apache Kafka to Apache Iceberg using Apache Flink SQL. The pipeline processes retail order data and stores it in Iceberg format on MinIO object storage.