3D Perception · Embodied AI · Efficient Edge Deployment
Multi-camera BEV architectures, monocular depth estimation, transformer-based 3D object detection, and multi-sensor fusion for autonomous driving. Deployed on NVIDIA Jetson Orin in sub-30 ms.
Vision-Language-Action models for grounded robot autonomy, knowledge distillation for edge VLA deployment, and closed-loop VLA policy validation in Isaac Sim and CARLA.
PTQ/QAT INT8–FP16 quantization, TensorRT/ONNX deployment, knowledge distillation, and LoRA/PEFT for real-time AI inference on NVIDIA Jetson Orin under tight power budgets.
Omer Tariq, M. Bilal, D. Han
Adaptive differential privacy + quantized federated learning for distributed edge intelligence.
IEEE IoT Journal — IF 8.9 2025
Omer Tariq, D. Han
298K-param transformer — 4.7× compute reduction over LSTM, real-time on embedded hardware.
IEEE IoT Journal — IF 8.9 2025
Omer Tariq, B. Dastagir, M. Bilal, D. Han
Sub-meter accuracy across diverse environments without environment-specific retraining.
IEEE IoT Journal — IF 8.9 2025
Neubility, Seoul (Aug 2025 – Present)
Multi-camera 3D perception; 8× inference speedup via TensorRT INT8/FP16; VLA policy validation in Isaac Sim & CARLA.
Monitra, UK (Feb 2025 – Sep 2025)
15% robustness gain + 8× throughput; Azure Kubernetes MLOps pipeline; diffusion-model data augmentation.
KAIST, South Korea (Aug 2021 – Jun 2025)
DeepILS, NanoMST, ConvXformer; multi-camera Vision Transformer for BEV detection at 25 FPS; FPGA SLAM accelerator.
NECOP, Pakistan (Apr 2019 – May 2022)
SoC/RTL verification (SystemVerilog/UVM); tape-out on first silicon pass for custom AI accelerator designs.
SUPARCO, Pakistan (Oct 2014 – Apr 2019)
Radiation-tolerant FPGA DSP systems; AI-accelerated image compression; Distinguished Service Award (2018).
Stanford University (Coursera)
LUMS, Pakistan
Georgia Tech (Coursera)
Udemy
EPFL (Coursera)
University of Washington (Coursera)