Naif Asswiel

AI Automation Engineer — CGI

• Fine-tuned an SLM with parameter-efficient techniques to generate business-oriented functional retrodocumentation from COBOL, evaluated via an automated LLM-as-judge benchmark.
• Designed and built CGI-Factory-AI, a spec-driven AI software factory for Claude Code: four chained plugins (scoping → architecture → design → assembly) that take a project from workshop notes to implementation-ready specs.
• Engineered human-gated orchestration with deterministic guardrails: validation scripts and full traceability ensure AI proposes, humans decide at every gate.
• Automated the handoff to delivery: the factory converges three contracts into a GitHub Spec Kit repository and auto-creates Linear tickets per feature and phase.
• Created Shapsha, a Claude Code plugin that runs the entire Python quality pipeline in one command: linting, two-pass security (bandit + LLM-native analysis), test generation, doc sync, conventional commits and PR.
• Designed its parallel fan-out sub-agent architecture: per-file isolated agents cut pipeline time to the slowest file instead of the sum, with zero silent failures.
• Built valido-bistro, an interactive workflow of five chained Claude Code skills that simulates a full restaurant ordering experience, from greeting to receipt.

Deep Learning Researcher — LIUM

• Applied and fine-tuned state-of-the-art models for audio-visual deepfake detection.
• Reproduced state-of-the-art results through large-scale model training.
• Leveraged temporal deepfake detection models to run experiments and assess their performance on benchmark datasets.

M.Sc. in Artificial Intelligence

B.Sc. in Computer Science

DELF B2 — French