Highly innovative. Good Analytical and Entrepreneurial skills that lead to find out newer ideas from existing systems and data.Highly adaptable. Continual self-learner. Enjoys working both as a team member and individual.
Leading an artificial intelligence team that is specialized in computer vision
In this role, I was part of the development team at KPST, where I worked on research and development projects that were frequently commissioned by ETRI."
Lead a team that is developing an Automated Tagging System server for object detection, face detection, pose estimation, and license plate detection.
The project involved creating a storyboard based on a user's sentence. The project utilized various computer vision skills including face detection, swapping, text detection and removal, inpainting, sketching images, captioning images, and retrieving similar sentences to the user's input.
As the main developer, created a program to help patent examiners detect illegal copying by retrieving similar images from a database of 3 million images, saving examiners time during the examination process.
Conducted research to compare the performance of TensorFlow and PyTorch for pose estimation and converting TensorFlow code to PyTorch and training a TensorFlow-trained dataset with PyTorch.
Worked as a developer on a project that aimed to detect food, requiring classification of more than a thousand classes and processing hundreds of thousands of images.
Primarily use Python, but have experience with C++ for projects where speed performance is critical. Have experience converting torch models to optimized models for deployment on mobile apps (tflite, PyTorch Mobile, TorchScript), web (WEBGL with tflite), and API servers (Triton, Kserve, Flask, Django, FastAPI).
Experienced in object detection, from small datasets with small objects to large datasets. Familiar with image retrieval and segmentation.
Not an expert in serving models, but have experience deploying AI models with inference server engines.
Capable of implementing cutting-edge AI models from research papers.