- Works at AI R&D Team, AI Division, AI Unit.- Obtained Ceritified Investment Manager certification from Korea Financial Investment Association. (March 2024)- Passed Financial Risk Manager (FRM) - part 1 exam (September 2024)Main responsibilities- Research on portfolio management using Mamba model with Deep Ensemble technique.- Developing market-preceding leading indicators using Machine Learning methods.- Research on Probabilistic Time Series Models.Secondary… Show more - Works at AI R&D Team, AI Division, AI Unit.- Obtained Ceritified Investment Manager certification from Korea Financial Investment Association. (March 2024)- Passed Financial Risk Manager (FRM) - part 1 exam (September 2024)Main responsibilities- Research on portfolio management using Mamba model with Deep Ensemble technique.- Developing market-preceding leading indicators using Machine Learning methods.- Research on Probabilistic Time Series Models.Secondary responsibilities- Implemented secured ChatGPT API inference service within company platform.- Automated fund management process of 8 million+ $ fundamental quant fund with Python.- Automated stock order distribution number decision with Machine Learning methods. Show less

- Works at AI R&D Team, AI Division, AI Unit.- Experimented Deep Learning Time Series models to asset data to estimate quantiles of future forecasts.- Experimented Deep Probabilistic Time Series Models to asset data to estimate predictive distribution.- Calculated VaR and Expected Shortfall with Deep Probabilistic Time Series Models' prediction.- Estimated covariance matrices with Deep Probabilistic Time Series Models.- Performed Mean-Variance Optimization with predictions… Show more - Works at AI R&D Team, AI Division, AI Unit.- Experimented Deep Learning Time Series models to asset data to estimate quantiles of future forecasts.- Experimented Deep Probabilistic Time Series Models to asset data to estimate predictive distribution.- Calculated VaR and Expected Shortfall with Deep Probabilistic Time Series Models' prediction.- Estimated covariance matrices with Deep Probabilistic Time Series Models.- Performed Mean-Variance Optimization with predictions and estimated covariance matrices. Show less

- Worked on Text Classification and Quantization project.- Designed domain-specific text preprocessing scheme.- Made a classification baseline for speech intention classification of Korean STT data. (73% Acc)- Experimented methods to mitigate Imbalanced Class, Label Noise and Open Set Learning problems.- Accelerated the inference with Quantization by ~1.7x- Accelerated the inference with Model Serialization (TorchScript and ONNX export) by ~1.83x.

- Worked at AI Research and Development Division, Republic of Korea Army Training and Doctrine Command.- Worked on Proof-of-Concepts of the AI applications to the military.- Challenged by limited compute resources and strict security rules.- Built and maintained AWS hosting server systems to utilize remote GPU compute servers. (Docker)- Managed a speech recognition system and a video synthesis system for military application.- PyTorch Contributions: Wishart Distribution… Show more - Worked at AI Research and Development Division, Republic of Korea Army Training and Doctrine Command.- Worked on Proof-of-Concepts of the AI applications to the military.- Challenged by limited compute resources and strict security rules.- Built and maintained AWS hosting server systems to utilize remote GPU compute servers. (Docker)- Managed a speech recognition system and a video synthesis system for military application.- PyTorch Contributions: Wishart Distribution, Constraints for distributions- Sponsored for 2022 Asian Conference on Machine Learning and Online Asian Machine Learning School Show less

- Participated research project ‘Physics-informed study and application of Artificial Intelligence’- Implemented proposed optimization algorithm related to Langevin dynamics. (TensorFlow)- Experimented on Image classification benchmarks.- Experimented the effect to Ensemble Method.

- Enhanced privacy in chatbot service (NLP, dialogue generation) (Huggingface, SpaCy, Docker)- Trained NLP models to detect personal information in the generated text.- Removed or transformed sensitive information to anonymize the generated text.- Deployed the model to real service and documented the project for the maintenance.- Completed MLOps Coursera Course in 1 week with the knowledge and experience from the internship.

- Trained video detection model (R-CNN and Faster R-CNN) for real-time human detection system.- Optimized the model with Quantization techniques.- Independent study on VAE, Gaussian Process, Bayesian Neural Networks variants (Pytorch)

- Participated in research project related to Business Intelligence- Collected and refined data from web to utilize formal data for the data analysis project.