Hi, It' Siwarat

As a member of the Text-to-Speech (TTS) team, I contributed to developing innovative TTS models for global languages while enhancing existing systems. My role encompassed the entire TTS workflow, from data exploration and collection to model development and optimization.

• Conducted comprehensive analysis of speech datasets, identifying key features for high-quality synthesis
• Contributed to the development of multi-lingual TTS systems, expanding language coverage
• Collaborated with cross-functional teams to integrate TTS solutions into various applications
• Designed and executed experiments to improve naturalness and expressiveness of synthesized speech

Working on developing a novel deep learning model for image segmentation

• Led development of automated car damage severity detection system, improving efficiency in automotive repair evaluation
• Implemented high-accuracy instance segmentation models for car part identification and damage detection
• Developed CNN and transformer-based car deformation severity classification model
• Deployed car severity prediction pipeline on AWS ECS
• Designed air-coil-condition classification model for HVAC systems, enhancing efficiency

Participated in an intensive AI bootcamp program, developing innovative machine learning solutions for diverse real-world challenges. Demonstrated versatility in applying AI techniques across multiple domains, from geospatial analysis to computer vision and natural language processing.

• Developed AutoML-based poverty rate estimation model using geospatial data from QGIS and GEE
• Implemented logo search model combining ViT-base autoencoder, OCR, and k-nearest neighbor algorithms
• Created YOLOv8x model for Liver Lesion Detection
• Designed long document retrieval pipeline using BERT and Elasticsearch
• Built image regression model for human motion prediction from ultra-wide-band signals