Atlanta, Georgia, United States
Adapted existing NWB‑to‑Mujoco pipeline from rat to mouse data, generating motion videos Reconstruct 3D arena from multi‑view videos to synthesize mouse’s first‑person perspective
Trained and evaluated CNN + LSTM and S4 models in 2D minigrid environments to solve mazes Integrated liquid ODE, joint training and VLA into Mamba State Space Model to improve Atari gameplay Refined RL policy training on world models for financial market prediction
Engineered full-stack platform with advanced image editing for 2D mechanical design Created web application to interact with 3D models and generate real-life images simplifying design process Implemented minimap system in Unity enabling object labeling and interaction in Meta XR simulator
Constructed React-based system for text- and voice-driven CAD model creation and editing Generated 6-view images from CAD models to support CAD-generative AI model training Developed VR headset-integrated Unity interface for stepwise CAD generation using gestures and voice
Established scraping pipeline to collect 500+ TikTok and Instagram Minecraft videos Applied Llama 3.2 model for video summarization and sentiment analysis Stored data with Databricks Spark SQL and set up RAG to retrieve gameplay insights