Haithem Chaabeni

Developed AI-powered content generation and optimization solutions using LLMs, Generative AI, prompt engineering, and structured outputs. Designed agentic workflows and RAG pipelines using LangGraph, LangChain, embeddings, semantic search, and FAISS.

Developed a full NLP pipeline for intelligent analysis, structuring, and matching of CVs with job offers using LLMs and semantic embeddings. Integrated anomaly and fraud detection modules to improve dataset reliability and recommendation quality.

Developed wheat disease detection models using YOLOv5/YOLOv8 and computer vision techniques. Deployed models on NVIDIA Jetson Nano with TensorRT and created Power BI dashboards for performance monitoring.

Built four deep learning models (EfficientNet, CNN, ResNet50, and U-Net) for lung cancer detection, covering nodule localization, benign/malignant classification, cancer stage classification, and cause verification. Trained and validated on multiple datasets including LUNA16, achieving 83-86% accuracy with low loss values.

Implemented diffusion-based generative AI for data augmentation, enhancing feature extraction. Engineered a multi-modal architecture combining mel-spectrogram transformations with ResNet-50 for audio classification and speech-to-text pipelines for NLP-based emotion analysis.

Pursuing an Engineering degree in Data Science & AI, with specialization in LLM systems and intelligent applications.