Duc Anh Ho

Built robust data pipelines for advanced instance segmentation and object detection models, using FastAPI for efficient REST APIs. Employed PostgreSQL for data storage, applying outlier detection and imputation for high-quality datasets. Utilized Kafka for microservices communication, connecting FastAPI services for seamless data flow. Designed and automated CI/CD pipelines using GitHub Action and ensuring smooth model deployment. Created optimized DB structures for efficient data retrieval and storage.

Developed ML-based building classification system utilizing Computer Vision techniques for efficient urban planning, customized UNet and DeepLabv3+ models to accurately identify building amenities from geospatial data. Researched large language model (LLM) and developed Text-to-Image synthesis projects using Transformer-based models, generating realistic images from textual descriptions. Collaborated with a team to preprocess data, fine-tune the model, and deploy a powerful tool of Computer vision and LLM