All Projects

VALORANT Round Outcome Prediction

Collaborated with NAIST to develop a transformer-based deep learning approach for predicting round outcomes using minimap video analysis. Achieved 80.55% accuracy and constructed a dataset from 1,376 tournament videos.

Python Deep Learning Computer Vision TimeSformer OpenCV

Texture Platform Automation

Developed an automated texture packaging system in Unity and Unreal Engine using Python and LLM integration for Bird Fab Studio. Reduced review failures to nearly zero and implemented automatic proofreading/translation.

Python LLM Automation Productivity Tool

LLM Bias in Algorithm Evolution

Researching LLM biases in heuristic algorithm generation using frameworks like ShinkaEvolve and OpenEvolve. Investigating why LLMs tend toward local optimization and avoid bold algorithmic modifications.

LLM Research Algorithms Heuristics Bias Analysis

Speech-to-Speech LLM Agent

Developed a comprehensive Speech-to-Speech LLM Agent system at CLINKS Corporation. Independently managed the project from initial design to implementation as a PoC.

Python LLM Speech-to-Speech AI Agent R&D

AI Agent Evaluation Framework

Researching and implementing standard methodologies for AI agent evaluation frameworks at Not A Hotel Inc. Focusing on LLM as a Judge approach for automated AI system assessment.

Python LLM AI Evaluation Research Automation