AI Engineer at iCog Labs

Contributed to the development and optimization of the Economic Attention Allocation (ECA N) system— a cognitive architecture component for intelligent resource management in AI systems while integrating symbolic AI principles for efficient computational resource management.Tested multiple core functionalities in C++ and Metta(the language of the cognitive architecture of OpenCog Hyperon of singularitynet). Contributed to NeuroGraph AI Assistant — an intelligent subgraph mining system that discovers frequent patterns in large-scale graph data (e.g., cryptocurrency transactions) using Graph Neural Networks (GNNs) and search algorithms (MCTS, Beam Search). Designed and implemented a motif interpreter powered by Gemini LLM to translate complex graph patterns into human-readable, expert-level explanations. Replaced brittle log-parsing logic with a structured JSON pipeline, enabling reliable LLM input and scalable interpretation. Integrated the interpreter with a web-based visualizer, creating an interactive chatbot interface for pattern analysis. Improved the Beam Search decoding strategy in the neural subgraph miner by fixing path reconstruction and scoring inconsistencies, enhancing pattern discovery accuracy and stability.