MSc Software Engineering (GL) · AI Researcher · Medical AI Engineer
Building neurosymbolic and multimodal AI systems at the intersection of clinical reasoning and edge deployment.
I hold a Master's degree in Software Engineering from Université Ibn Khaldoun, Tiaret, supervised by Prof. Tahar Zioual. My thesis - NeuroDetect Lite - is a lightweight neurosymbolic clinical decision support system for early Alzheimer's detection, co-authored with Kheireddine Belghitar. It received an Excellent mark at defense.
Before CS, I spent two years in medical school - which shapes how I think about clinical AI: precision matters, and explainability isn't optional.
I also presented ClinReasonViz at IEEE PAIS'26 with Prof. Abdelkader Ouared.
My current focus is two-track:
- 🔬 Extending NeuroDetect Lite into a peer-reviewed journal paper (VLM + QLoRA on OASIS-3, targeting Computers in Biology and Medicine)
- 💼 Securing remote AI/ML engineering roles - open to full-time, contract, and freelance work
- Neurosymbolic MMSE Gate: 7 DSM-5 clinical rules boosted ensemble accuracy from 55.68% → 73.86% (McNemar p = 0.003) - the core contribution of the thesis
- LightAlzNet: 0.63M params · 1.93MB INT8 · 10ms CPU inference — outperformed larger ImageNet-pretrained models on structural MRI
- Selective quantization: Kept Linear layers in FP32, quantized Conv2d only — reduced INT8 F1 degradation to ~0.26%
- Live demo: neuro-detect-lite.vercel.app
PyTorch · ONNX Runtime · TorchIO · scikit-learn · QLoRA · Hugging Face Transformers · FastAPI
Tauri · Vercel · Android (Kotlin)
- 📄 VLM paper: LFM2.5-VL-450M + QLoRA on OASIS-3 - generative diagnostics extension of the NeuroDetect Lite baseline, targeting Elsevier Computers in Biology and Medicine
- 💼 Remote AI/ML roles: Open to full-time and contract positions in medical AI, agentic systems, and edge deployment
- 🛠️ Freelance: Available on Upwork for AI/ML engineering and full-stack projects
