Hussein Ertimi

I’m an Electrical & Electronic Engineering graduate (First Class Honors, University of Tripoli), currently an AI & Software Engineer at the Central Bank of Libya and an incoming MASc student in AI at the University of Guelph.

My research centers on differentiable logic-gate networks: I rebuilt and reproduced a NeurIPS 2024 method from scratch, extended it to semantic segmentation with novel logic-based upsampling operators, deployed a trained network on an FPGA, and am now preparing the work as a first-author paper. I’ve also done independent mechanistic-interpretability work, isolating a steering vector for sycophancy in a reasoning model, and built NLP systems including an Arabic retrieval-augmented QA toolkit.

Looking forward, my interests center on foundation models and LLMs, AI safety, alignment, and interpretability, and safe, efficient, hardware-aware AI. I’ve built working knowledge across these—transformer internals, activation steering, how large models are trained and evaluated—and I’m deliberately keeping the range wide before committing my research to one of them.

At the bank I ship production systems end to end: AI document processing (OCR→LLM extraction), backend services over multiple production databases, and the system design, data engineering, and deployment work behind them.

Skills

• Programming: Python, C/C++
• Machine Learning: PyTorch, scikit-learn, Hugging Face
• Backend & System Design: FastAPI, microservices, APIs, DBs, Docker
• Data Engineering: SQL, pandas, NumPy
• Hardware & FPGA: Verilog, Digital Logic Design, Quartus Prime, ModelSim/Questa
• DevOps & Tools: Azure, Docker Compose, Linux, Git

Interests

• Foundation Models / LLMs
• AI Safety, Alignment and Interpretability
• Efficient & Hardware-Aware ML
• Differentiable Logic Networks