AliAkarma
I build agentic AI systems that know their own limits. My work focuses on the gap between autonomous capability and institutional accountability — designing architectures where AI agents can be stopped, audited, and corrected when they behave unexpectedly. I'm a 4th-year IT student at the Islamic University of Madinah and have published 19 peer-reviewed papers and manuscripts on AI governance, adversarial robustness, and constrained multi-agent systems. Read background →
I study how to make autonomous AI systems fail safely: designing governance architectures that prevent unintended actions before they propagate through real-world infrastructure.
Current Research Frontier
Autonomous Safety Governance
Safety-Critical Multi-Agent Systems
"Investigating cryptographic trust-anchors and constrained reasoning for large-scale agentic deployments."
Why This
Research
Matters
I study how to make autonomous AI systems fail safely: designing governance architectures that prevent unintended actions before they propagate through real-world infrastructure.
My work addresses the alignment problem in deployed agentic systems — exploring how we can build autonomous pipelines that remain safe and governable when exposed to adversarial inputs, distributional shift, or misaligned incentives. I approach this through the intersection of safety engineering, formal governance frameworks, and empirical failure-mode analysis.
Research Landscape
Recent Publications
Privacy Leakage in Federated Learning: Gradient-Based Client Identity Inference and Defenses for Inertial Sensing in Vehicular Edge Networks
Ali Akarma, Toqeer Ali Syed, Muhammad Khan, Qurat-ul-ain Mastoi, Adeel Ahmad
Agentic AI for Climate-Resilient Cities: A PRISMA-Guided Review and Digital Twin Framework
Toqeer Ali Syed, Ali Akarma, Muhammad Tayyab Naqash, Danial Hameed, Shahid Kamal, Antonio Formisano
Agentic AI for Inclusive Assistive Ecosystems: Architecture, Governance, and Personalized Support for People with Disabilities
Ali Akarma, Toqeer Ali Syed, Hammad Muneer, Danial Hameed
Featured Research Systems
LagrangianCTDE — Constrained MARL for Disaster Response
Risk-aware multi-agent reinforcement learning framework coordinating Storm, Flood, and Evacuation agents under Lagrangian safety constraints, achieving 81.5 reward with only 2.3% safety violations across six baselines.
Agentic AI-Enhanced Digital Twin — Smart City Infrastructure
Physics-grounded simulation framework evaluating rule-based, digital twin, and agentic AI monitoring architectures for smart city civil infrastructure, with blockchain-anchored audit trails and Kalman-filtered state estimation.
FinNutriAgent — Household Budget & Nutrition Optimizer
Open agentic AI framework jointly optimizing household meal planning and financial budgets under nutritional, cultural, and economic constraints using MILP and LLM orchestration across multi-store price data.
Latest Research Notes
Privacy Leakage in Federated Learning: Client Identity Inference and Defenses for Inertial Sensing in Vehicular Networks
Federated learning is widely heralded as privacy-preserving because raw sensor data never leaves the edge. This paper presented at IEEE VTC 2026 reveals that undefended weight deltas allow an honest-but-curious server to identify clients with near-perfect accuracy (≈1.000), and formulates rigorous clip-then-noise and ensemble defenses with formal (ε, δ)-DP guarantees.
Agentic AI for Inclusive Assistive Ecosystems: Architecture, Governance, and Personalized Support for People with Disabilities
Current assistive technologies often operate as fragmented, reactive tools. This book chapter introduces an agentic AI ecosystem combining multi-agent coordination, privacy-preserving governance, and personalized daily routines to empower individuals with disabilities with greater independence.
Recent Milestones
Presented paper: Privacy Leakage in Federated Learning: Gradient-Based Client Identity Inference and Defenses for Inertial Sensing in Vehicular Edge Networks — IEEE VTC 2026
New paper: Agentic AI for Climate-Resilient Cities: A PRISMA-Guided Review and Digital Twin Framework — MDPI Sustainability
New chapter: Agentic AI for Inclusive Assistive Ecosystems: Architecture, Governance, and Personalized Support for People with Disabilities — IGI Global
Building Safety-Aligned Autonomous AI
Currently open to research collaborations, academic exchanges, and graduate opportunities in AI safety, multi-agent systems, and trustworthy machine learning.