Graph ML
Graph Intelligence
Ontology-aware heterogeneous GNNs for incomplete archaeological ceramics.
98.33% accuracy
← Back to selected work
Project archive
All Projects
AI research, retrieval systems, computer vision, recommendations and product engineering.
LLM / RAG
RAG Benchmark
Benchmarking sparse vs dense retrieval and local LLM grounding.
100 test queries
Computer Vision
RunRightAI
Video-based sprint velocity estimation using YOLOv8, MediaPipe, and homography calibration.
5 Markers planar calibration
Recommendation
Adaptive Recommendations
Heterogeneous GNNs with reinforcement learning.
Personalized ranking
Software
Body Force
Gym management system with biometric access control.
Member tracking