Laith Benmabrouk

AI Engineer specialized in Python-based AI systems, Machine Learning, and LLM-driven automation. Experienced in designing scalable AI architectures including conversational agents, Retrieval-Augmented Generation (RAG) systems, and intelligent document processing pipelines. Proven track record of building production-ready AI solutions that automate complex enterprise workflows and reduce manual operations by over 70%.

Strong expertise in backend AI systems, data engineering, API integrations, and cloud deployment.

AI & Software Engineer at AVOCarbon Group (2025-05 – Present)

Designed and deployed production-grade AI systems and automation agents using Python and LLM APIs to orchestrate operational workflows across supply chain, sales, and engineering teams.

• Designed and deployed production-grade AI systems and automation agents using Python and LLM APIs to orchestrate operational workflows across supply chain, sales, and engineering teams.
• Architected and implemented an AI-powered EDI automation pipeline including email ingestion, regex-based document parsing, ETL transformation of unstructured order documents, and automated insertion into PostgreSQL databases via REST APIs, eliminating manual EDI processing.
• Developed a Retrieval-Augmented Generation (RAG) architecture integrating embeddings, vector databases, and semantic search (LangChain / HuggingFace) to enable contextual retrieval across enterprise knowledge bases and improve information accessibility.
• Built conversational AI agents integrating LLM APIs with enterprise databases and backend services to enable natural-language querying of operational data.
• Engineered a technical drawing analysis and costing engine combining OCR-based document digitization (PaddleOCR), LLM-driven extraction of engineering specifications from drawings, and rule-based pricing models to generate automated cost estimations.
• Implemented advanced prompt engineering, memory management, document intelligence, and semantic retrieval techniques to improve response accuracy, contextual relevance while reducing hallucinations in LLM systems.
• Performed model evaluation using precision/recall metrics and hyperparameter tuning to improve prediction accuracy.
• Worked in Agile teams collaborating with data engineers and product stakeholders.
• Business Impact: Automated critical workflows saving 200+ hours/month, reducing operational errors by 90%, and accelerating internal processing times by 70%.

AI & Data Engineer at STECC (2024-07 – 2025-04)

Built document intelligence pipelines combining OCR and NLP techniques to convert engineering documents and technical files into structured datasets for downstream analysis.

• Built document intelligence pipelines combining OCR and NLP techniques to convert engineering documents and technical files into structured datasets for downstream analysis.
• Applied data extraction, transformation, and validation techniques using Pandas and NumPy to prepare datasets for machine learning and operational reporting.
• Contributed to AI-assisted engineering tools capable of analyzing technical documentation and extracting structured specifications to support feasibility and production analysis.
• Collaborated in an agile development environment, iterating rapidly on prototypes and integrating feedback from engineering and operations teams.

End-of-Study Internship at Draxelmaier (2024-02 – 2024-06)

Designed and developed a WPF desktop application to automate production data collection and visualization.

• Designed and developed a WPF desktop application to automate production data collection and visualization.
• Integrated real-time analytics dashboards and reporting modules to support faster operational decision-making.
• Implemented data validation and integrity checks, reducing inconsistencies.
• Designed intuitive UI components focused on usability and efficiency.
• Produced technical documentation for maintainability and scalability.

Advanced Training Internship at Artibedded (2023-06 – 2023-08)

Collected, labeled, and augmented computer vision datasets to improve training robustness and model generalization.

• Collected, labeled, and augmented computer vision datasets to improve training robustness and model generalization.
• Trained and optimized deep learning computer vision models using PyTorch and TensorFlow for plant identification and health analysis.
• Integrated APIs for species classification and automated health assessment.
• Implemented dimension measurement for plant growth tracking.

Industrial Computer Science Engineering Degree in Industrial Computer Science Engineering – National School of Electronics and Telecommunication of Sfax (ENET'Com) (2021 – 2024)

Pre-Engineering Degree in Pre-Engineering – Preparatory Institute for Engineering Studies of Nabeul (2019 – 2021)