Modern AI systems are getting more powerful and more centralized at the same time. That creates real problems: privacy risks, single points of failure, and data that organizations cannot share. I build distributed AI frameworks that let multiple parties train models together without exposing raw data. My work covers healthcare, cybersecurity, and environmental monitoring.
Hospitals, banks, IoT networks, and mobile devices all hold sensitive data that AI needs. The problem is that privacy laws, competition, and security risks make it impossible to pool that data in one place. I build tools that let AI models train across separate data silos without moving or exposing any raw data.
I also study the security of decentralized systems themselves. Smart contracts and federated learning protocols have real vulnerabilities that attackers exploit. I build detection and defense tools grounded in representation learning and multimodal analysis to protect these systems.
Both research areas share a core principle: test under realistic adversarial conditions, not just clean benchmarks. Systems built with a clear threat model in mind are more secure and generalize better. My results consistently show this.
I build federated learning frameworks that train models across distributed data owners without centralizing any sensitive data. The main failure modes I address are backdoor attacks, heterogeneous data distributions, and communication bottlenecks that block real-world deployment.
RBBD reduces backdoor attack success below 15% while keeping model utility intact
Federated deforestation detection from satellite imagery reaches 0.85 IoU without sharing raw data
Aggregation strategies for non-IID medical and IoT data distributions
Related Publications
RBBD: A Representation-Based Framework for Edge-Case Backdoor Defense in Federated Learning· 2025 IEEE 7th International Conference on Trust, Privacy and Security in Intelligent Systems
Privacy Meets Conservation: Federated Learning's Revolution in Deforestation Detection · 2025 IEEE World AI IoT Congress
Smart contracts hold billions of dollars in value but break under subtle code vulnerabilities. I build multimodal tools that fuse source code, opcode, and bytecode to catch vulnerabilities that single-modality approaches miss.
VulnFusion: multimodal transformer for smart contract vulnerability detection
QuadraCode AI fuses source code, opcode, and bytecode representations (8 citations, ICCCN 2024)
Blockchain-IoT framework for supply chain transparency and traceability
Related Publications
VulnFusion: Exploiting Multimodal Representations for Advanced Smart Contract Vulnerability Detection · 2024 6th International Conference on Blockchain Computing and Applications
QuadraCode AI: Smart Contract Vulnerability Detection with Multimodal Representation · 2024 33rd International Conference on Computer Communications and Networks
A Blockchain and IoT-Enabled Framework for Ethical and Secure Coffee Supply Chains· Future Internet 17
I apply machine learning to safety-critical domains where reliability directly affects patient outcomes. My work covers brain tumor segmentation, Alzheimer's classification, ICU outcome prediction, cervical cancer detection, and physiological stress analysis.
ICU length-of-stay prediction from MIMIC-III (27 citations, most cited work)
fNIRS mental workload classification with CNN and ML ensembles (12 citations)
ACGAN-augmented brain tumor segmentation with independent validation
Scalable multimodal cervical cancer detection pipeline
Related Publications
Prediction of Length-of-Stay at ICU Using Machine Learning based on MIMIC-III Database · 2023 IEEE Conference on Artificial Intelligence
Mental Workload Classification from fNIRS Signals by Leveraging Machine Learning · 2023 IEEE Signal Processing in Medicine and Biology Symposium
Improved Classification of Alzheimer's Disease with Convolutional Neural Networks · 2023 IEEE Signal Processing in Medicine and Biology Symposium
Brain Tumor Segmentation and Classification Using ACGAN with U-Net and Independent CNN-Based Validation· 2024 IEEE Signal Processing in Medicine and Biology Symposium
Scalable Multimodal Machine Learning for Cervical Cancer Detection · 2024 IEEE World AI IoT Congress
Personalized Stress Detection from Chest-Worn Sensors by Leveraging Machine Learning · 2024 International Conference on Electrical, Computer and Energy Technologies
Scalable AI needs scalable infrastructure. I benchmark distributed data pipelines, optimize parallelization for preprocessing bottlenecks, and build cloud-native ML systems with measurable throughput and latency gains.
MPI4Py preprocessing optimization, 36 citations, published in Machine Learning with Applications
Apache Spark vs. Airflow benchmarking for batch ML workloads
Serverless AWS fraud detection system, 94.9% accuracy, low latency
Related Publications
Minimization of High Computational Cost in Data Preprocessing with MPI4Py · Machine Learning with Applications
A Big Data Optimization Comparison using Apache Spark and Apache Airflow · 2025 IEEE 15th Annual Computing and Communication Workshop
Federated learning and multi-agent systems are strategic environments. Participants can act selfishly, collude, or drop out to maximize their own gain. I apply mechanism design, Nash equilibria, and cooperative game theory to build distributed AI systems where honest participation is the rational choice.
Models FL participation as a strategic game under data heterogeneity and free-rider incentives
Mechanism design for incentive-compatible aggregation in cross-silo federated settings
Game-theoretic analysis of adversarial behavior in blockchain consensus protocols
Cooperative game theory for fair reward allocation among federated learning contributors
Applying federated learning to large language and vision-language models, where communication costs and data heterogeneity are far greater than in standard FL.
Building tighter, composable privacy accounting methods that give real guarantees for deployed federated systems without the accuracy loss of standard DP-SGD.
Combining static analysis, fuzzing, and learned representations to build audit pipelines that catch new vulnerability classes as Solidity and the EVM evolve.
Using federated multi-source remote sensing for climate and biodiversity monitoring, enabling global environmental models without centralizing sovereign satellite data.