PhD · AI / Computer Vision Researcher

Ahmed Ghali

I build efficient deep-learning systems for vision — from 1.58-bit Vision Transformers for edge devices to applied digital-twin and full-stack platforms.

01 / About

I'm a PhD researcher and university lecturer working at the intersection of machine learning and real-world engineering. My research focuses on efficient face-recognition architectures, few-shot learning, and low-bit Vision Transformers for edge deployment.

Alongside research, I ship production software (Python, Django) and founded AIOpenCamp, an open AI-education platform for the Arabic-speaking world. I'm currently open to postdoctoral and research opportunities — internationally and remote.

PhD, Computer ScienceSimulation & AI · 2025
University LecturerSince 2015 · Ouargla, Algeria
FounderAIOpenCamp
LanguagesArabic · French · English
02 / Research focus

What I work on

Computer Vision Deep Learning Face Recognition Efficient / Edge AI Vision Transformers Few-shot Learning Digital Twins Optimization & Metaheuristics
03 / Selected publications
2025
BitFace: A 1.58-bit Vision Transformer for Energy-Efficient Face Recognition on Edge DevicesLatest
SSRN preprint · A. Ghali, A. Khaldi, T. Ghrib
2023
Few-Shot Learning CNN Optimized Using Combined 2D-DWT Injection and Evolutionary Optimization Algorithms for Human Face Recognition
Int. J. of Wavelets, Multiresolution and Information Processing · A. Ghali et al.
2014
Improving the Recognition of Faces using PCA and SVM Optimized by DWT
Int. J. of Computer Applications · G. Ahmed, B. Mohamed

View full publication list on Google Scholar →

04 / Projects

Things I've built & led

Digital Twin for ESP Systems

Supervised Master's research on an AI-driven digital twin for predictive maintenance and optimization of electrical submersible pumps.

Co-supervised with an engineer from SLB (Schlumberger)

BitFace

Energy-efficient face recognition built on a 1.58-bit (ternary) Vision Transformer, designed for low-power edge deployment.

PyTorch · Vision Transformers · Edge AI

Smart-Univ

Full-stack Django platform for university management — admissions, records, and administration.

Serving 5,000+ students & staff

AIOpenCamp

Open platform I founded to teach applied AI to the Arabic tech community — courses, articles, projects, and tools.

18 courses · 160+ articles
05 / Writing in English
Article

8 LLM Cost Optimization Techniques Every AI Engineer Should Know

A practical guide to cutting LLM spend by 50–80% — prompt compression, semantic caching, model routing, cascading, and continuous cost monitoring.

Read on AIOpenCamp →
06 / Contact

Open to research roles abroad

I'm looking for postdoctoral and research opportunities in computer vision, efficient deep learning, and applied AI — open to relocation and remote work. The fastest way to reach me is email.