Research & innovation
We take ideas from the research frontier and engineer them into systems that hold up in production.
The work below was led or co-authored by our co-founder with academic and institutional collaborators. It is individual research, not proprietary BodhixAI intellectual property. Each item is marked as published, accepted (forthcoming), or ongoing.
Fusing graph structure with flow features, and calibrating every alert with conformal prediction.
SAGE-IDS: Graph-centric Intrusion Detection with Structural Surprise and Conformal Risk for the IoX
A head-to-head study of classical and quantum models — awarded Outstanding Student Paper.
Classical vs. Quantum Approaches for Predicting Galaxy Velocity Dispersion in MaNGA Galaxies
Collaborative intrusion detection across industrial sites without sharing raw traffic.
FedGuard-IIoT: A Scalable Federated Learning Architecture for Network-IDS in the Industrial IoT
Federated learning with blockchain provenance for vehicular networks.
FL-DECO-BC: A Privacy and Provenance Preserving, Provably Secure FL Framework for VANETs
Teaching detectors easy-to-hard, and explaining what they learned.
Curriculum Learning with Image Transformation and XAI for Improved Network Intrusion Detection
Exploring quantum-inspired context modelling for compact, fast intrusion detection.
Quantum-Contextual Nano-LCM: A Lightweight Transformer for High-Performance IoT NIDS
Using physical and system constraints to make security models more robust and trustworthy.
Physics-aware / constraint-aware intelligent systems for network security in IoX
Evaluating whether agents can reason through real operational incidents, step by step.
ORBIT: benchmarking multistep operational reasoning in ITOps