Systems, experiments, and tools.
Multi-agent web text collector with planner, explorer, harvester, and auditor agents plus LLM-as-a-Judge quality control.
From diagrams to code — turning architecture diagrams into working scaffolding with LLMs.
Hybrid MCTS and Cross-Entropy planning framework for adaptive bathymetric mapping with autonomous underwater vehicles.
LLM agents from zero to hero — design patterns for tool use, planning, and memory.
Closed-loop autonomous navigation in ROS/Gazebo combining kinodynamic planning with reactive obstacle avoidance.
Deep learning for Parkinson's stratification across multi-omics data, with SHAP/LIME biomarker discovery.
Spatial distribution analysis of galaxies with significant GPU speedup — pure C, OpenMP, and CUDA implementations.