Freelancer profile translated to English.

Description

Roland BAZ — LLM Engineer · Data Scientist · LangChain · GPT · AI / ML 🚀

With a strong background in Artificial Intelligence, NLP, and Machine Learning (Paris Dauphine— one of the top engineering schools in France), I combine LLM expertise, Data Science skills, and software engineering to create high-performing and truly useful systems. Additionally, I am currently an AI Engineer with approximately 3+ years of experience in the field.

My services cover all GenAI & Data Science needs:

🌟 GENAI / LLM ENGINEERING

Do you want to:

✅ Build a chatbot or assistant with LangChain, OpenAI API, or Anthropic?

✅ Create a multi-agent system capable of using tools (web search, APIs, databases, vectorstores...)?

✅ Set up an RAG pipeline (Qdrant, Weaviate, FAISS...)?

✅ Optimize your chatbot for better accuracy, speed, and cost reduction?

✅ Fix bugs, hallucinations, or performance issues in your LLM system?

I will guide you from idea to deployment.

📊 DATA SCIENCE & MACHINE LEARNING

Need help with:

✔️ Developing predictive models: classification, regression, forecasting, anomaly detection?

✔️ Working on NLP projects (sentiment analysis, classification, fine-tuning)?

✔️ Implementing explainability (SHAP, LIME) and evaluation tools?

✔️ Deploying your models via Docker, FastAPI, CI/CD?

✔️ Transforming raw data into insights, dashboards, or automated systems?

I deliver fully operational, reliable, and interpretable ML solutions.

💬 Let's build your AI project

Whether you need an intelligent chatbot, a multi-agent system, a data science model, or a complete Generative AI product, I can help you design, optimize, and deploy it.

Ready to get started?

Contact me on Malt and let's transform your idea into a powerful AI solution. 🔥

Industry field of expertise

Languages

French
Native or bilingual
English
Fluent
Arabic
Native or bilingual
Armenian
Basic
Swedish
Conversational

Workplace preferences

Can work on-site

Paris (up to 50km)

Lutessa – Cabinet de Conseil
AI Engineer
CONSULTING AND AUDITS
April 2025 - Today (1 year and 2 months)
Paris, France
• Led the design and deployment of two AI systems in production for pharmaceutical clients:
1. Pharmacovigilance agent automation for Atessia, detecting drug safety signals in scientific literature, improving productivity by 70%; and
2. Document compliance validation for Novartis, automating the detection of pharmaceutical information and reducing review time from 4 hours to 2 minutes per document
AI Agents LangGraph AWS Docker Langchain
Recherche & Développement - Collaboration à Distance
Data Scientist
RESEARCH
January 2024 - December 2024 (11 months)
Paris, France
• Led two parallel Machine Learning research projects (Health & Economics) focused on benchmarking and explainability
• Designed a comprehensive ML pipeline comparing ensemble methods (XGBoost, Random Forest) against classical models (SVM, KNN) to optimize prediction performance in both domains
• Developed a detection system (lung cancer on over 4,500 patient records), using SMOTE technique to handle class imbalance and achieving an AUC of 0.91 (88% Precision)
• Rigorously applied the SHAP framework to models to identify critical biomarkers and quantify variable impact, ensuring clinical and economic interpretability of results

Technical Stack: Python, XGBoost, Random Forest, SVM, KNN, Logistic Regression, Scikit-learn, Pandas, NumPy, SHAP, Imbalanced-learn (SMOTE), Matplotlib/Seaborn
XGBoost Random Forest Scikit-learn Matplotlib Logistic Regression
LAMSADE – Laboratoire de Recherche
AI Researcher
RESEARCH
July 2023 - January 2024 (6 months)
Paris, France
• Developed an AI system for automated dialogue generation for a Swiss university, achieving a 50% improvement in coherence compared to the base GPT-3.5 model [Paper Link]
• Architected an end-to-end LLM pipeline, including fine-tuning LLaMA 3 (8B) and Mistral (7B) via 4-bit/8-bit QLoRA quantization for efficient GPU training on a theatrical corpus
• Implemented an RAG system with hybrid search (semantic + keyword) and chunking to ground models in the specific styles of playwrights
• Enhanced performance through prompt engineering (Chain-of-Thought, few-shot) and validation by theater experts
• Managed on-premise deployment (Linux / A6000 GPU) and developed the interactive interface using Flask for the backend

Technical Stack: Python, PyTorch, HuggingFace Transformers, QLoRA, LLaMA 3, Mistral, LangChain, RAG, Vector Databases, Flask, HTML/CSS/JavaScript, Linux, Git
Git Linux SQL Langchain Streamlit