Atif Abid

P.E.

AI for Physical Systems

PhD Candidate · Robotics & AI Engineering, SMME NUST
Visiting Faculty · Dept. of Mechatronics Engineering, College of EME, NUST

R&D Engineer Researcher Educator

Research: Edge AI & IoT • Precision Agriculture • Computer Vision & Robotic Perception • Scientific ML (PINNs & GNNs) • Generative Design

I build AI that respects physics. My work spans neuro-symbolic pipelines (Text2CAD), GNN surrogates for FEA, and Edge AI perception models designed for deployment on constrained hardware—where accuracy and real-world deployability must coexist.

My PhD at NUST focuses on Edge AI and IoT for Precision Agriculture, developing object detection systems that run reliably on resource-limited field devices.

🌱 Recently wrapped my first undergraduate course — Fundamentals of Programming (C++) — where students built autonomous line-following robots as their final project. The project videos got a lot of appreciation. Designing for constrained hardware starts in the classroom.

News

Dec 2025 Started as Research Visiting Faculty, Dept. of Mechatronics Engineering, College of EME, NUST.

Dec 2025 Presented poster on AI and IoT in Precision Agriculture at IEEE ICODT2 (SINES, NUST).

Nov 2025 Released preprint: Review of Last Mile Urban Logistics on Zenodo.

Jan 2025 Started PhD in Robotics & AI at NUST.

Publications & Preprints

Low Cost Edge-Based Person Detection for IoT Surveillance Systems Under Review

Atif Abid, Dr Danish Hussain

Submitted to IEEE Conference, 2026

Deep Learning for Wheat Disease Detection: A Comprehensive Review In Preparation

Atif Abid

2026

Review of Last Mile Urban Logistics: Policy & Safety

Atif Abid

Preprint, 2025 · Under Review

[PDF] [DOI / Zenodo]

Review of High-Impact Works: AI & UAVs for Precision Agriculture Poster

Atif Abid, Dr K. Faraz, Dr Shahbaz K.

IEEE ICODT2, Dec 2025

[View Poster PDF]

Selected Projects

Prompt2CAD — Neuro-Symbolic Generative Design [Code] [Demo] LLM CAD

Text-to-CAD pipeline that translates natural language into executable SolidWorks geometric feature trees, guaranteeing 100% geometric validity. Bridges LLM reasoning with physical design constraints.

Edge-Based Person Detection System [Demo] IoT Edge AI

Real-time surveillance node for constrained IoT hardware. Optimized lightweight detection models for person and object detection in agricultural settings. Paper under submission.

Reproducibility Study: SCS-YOLO for Precision Agriculture [Code] Computer Vision PyTorch

Re-implemented the SCS-YOLO architecture (originally in Computers and Electronics in Agriculture) on Google Colab. Validated on open-source pest/crop datasets, achieving mAP results consistent with the original authors.

Local LLM Inference Benchmarks [Code] [Data] LLM Benchmarking

Rigorous benchmarking of Llama 3, Qwen, and gptoss20B on consumer hardware (RTX 40-series). Evaluated vLLM, Ollama, and TensorRT-LLM to identify optimal local inference setups.

Edge AI: Model Compression for Agriculture [Code] TFLite Compression

3-stage compression pipeline (Pruning → Quantization → TFLite export) for RNN-based precision agriculture models, enabling real-time inference on legacy Android devices.