Samidur Rahman

High-performance chaotic dynamical system visualization achieving 100+ FPS with 15,000+ trajectory points. Represents 30-100× performance improvement over Python prototype through native C++ OpenGL implementation.

• Implemented 4th-order Runge-Kutta numerical integration for Lorenz equations
• Discovered critical numerical precision requirements with compiler optimizations
• Eliminated CPU-side rendering bottlenecks through GPU-accelerated rendering

ROS2-based autonomous navigation system with sensor fusion and path planning algorithms, implementing SLAM mapping and obstacle avoidance with 95% navigation accuracy.

• Developed C++ nodes for sensor fusion and Python path planning algorithms
• Reduced control loop latency by 40% through optimized message passing
• Integrated real-time sensor data processing with low-latency motor control

Built and deployed a quantised CNN→SNN epilepsy-detection pipeline on physical BrainChip AKD1000 silicon. Under chronological leave-one-patient-out (LOPO) protocol achieved 0.59 mean event sensitivity, 4.00 FP/hr, 0.866 specificity at ~0.906 µJ/inference chip-level energy cost.

• Hardware: Raspberry Pi 5 + AKD1000 M.2 neuromorphic processor (PCIe M.2 HAT+)
• Model: 137,826 parameters, 163 KB on-chip footprint, 4th-order Runge-Kutta integration
• Cross-patient results: 0.59 sensitivity, 0.866 specificity on held-out patient (LOPO evaluation)
• Discovered and documented 23 undocumented AKD1000 v1 hardware constraints via systematic probing
• Spatio-temporal CNN v2 with (9,7) kernels; chronological CHB-MIT splits, SMOTE balancing, per-tensor w4a4 quantisation
• Research: Cardiff Agile CPS Lab, supervised by Dr. Nick Pham

High-efficiency automated fact-checker using novel counter-claim generation framework with Semantic Role Labeling and dependency parsing.

• Fine-tuned BioPubMedBERT classifier achieving 65%+ accuracy and 70%+ supported-claim recall
• Optimized for deployment on consumer-grade hardware
• Created robust, balanced dataset for medical claim verification