TOFreelancer profile translated to English.
Back to original languageAbout Tarik
Data Scientist & ML Engineer with over 9 years of experience, I combine expertise in advanced machine learning and generative AI technologies (RAG, LLMs) to create high-value solutions. My technical expertise covers the entire data value chain - from developing robust pipelines on AWS to implementing intelligent conversational systems, including deploying ML models in production. Proficient in Python, SQL, Spark ecosystems and modern frameworks (LangChain, Langraph, FastAPI), I apply MLOps and software engineering best practices to ensure quality and scalability. My ability to translate business needs into innovative technical solutions, combined with constant technological monitoring, allows me to bring strategic value to data-driven projects.
French
Native or bilingual
English
Fluent
Can work on-site
Paris (up to 50km)
Experience
- BPCEML EngineerBANKING AND INSURANCEOctober 2025 - Today (10 months)Paris, FranceDesign and development of an end-to-end RAG pipeline for natural language querying of a large and heterogeneous document base (PDF, Word). The system relies on a hierarchical indexing architecture and hybrid dense+sparse retrieval to maximize the accuracy of generated responses, within an on-premise banking environment ensuring data sovereignty.
- JAAGML Engineer / AI DevCONSULTING AND AUDITSMarch 2025 - Today (1 year and 5 months)Paris, FranceContext:Design and development of a complete invoice processing automation solution for SMEs and accounting firms, using local artificial intelligence (Mistral via Ollama) and advanced OCR to transform PDF documents into usable structured data. The platform ensures total confidentiality with 100% local processing, reducing processing time by 90% (20 min → 2 min per invoice) with accuracy >95% on standard documents.Achievements:• Design of a modular architecture with clear separation of responsibilities (extraction, AI processing, storage)• Development of flexible data models with Pydantic for validation and complete metadata.• Implementation of dual extraction pipelines (PyPDF2 for digital PDFs, Tesseract OCR for scanned documents) with automatic document type detection and intelligent routing to the optimal method.• Configuration and integration of Ollama/Mistral for local document processing, development of JSON parsing strategies with fallbacks to ensure extraction even with malformed LLM responses.• Design and implementation of a PostgreSQL storage layer with optimized JSON indexing, multi-criteria search system (date, amount, type, content), and custom Excel/JSON export functionalities.• Development of a Streamlit demo interface with JSON viewer to validate UX/UI concepts and provide detailed specifications to the front-end dev for production implementation.• Writing detailed project documentation.Technical environment: Python, Streamlit, PostgreSQL, Ollama, Mistral, Tesseract OCR, PyPDF2, Pydantic, Pandas, NumPy, FastAPI, Git, Shell scripting, SQL, JSON Processing, Logging.
- Disneyland ParisML Engineer / AI DeveloperENTERTAINMENT AND LEISUREFebruary 2024 - February 2025 (1 year)Paris, FranceProject: Development and deployment of a recommendation algorithmContext:• Personalized recommendation system: Development and deployment of a personalized recommendation engine for Disneyland Paris, using predictive analysis and machine learning to suggest hotels tailored to visitor preferences. The system leverages enriched customer data and a cloud architecture to optimize conversions and user experience.• Intelligent document assistant (RAG): Design of a conversational system allowing internal teams to naturally query the technical documentation of applications developed by Disney, using RAG and LLM technologies to transform thousands of pages of documentation into a virtual assistant capable of accurately answering technical questions, thereby accelerating problem-solving.Achievements:◦ Collaboration with Product Owners to define objectives and evaluate existing models with Python and SageMaker◦ Development of extraction pipelines from Snowflake and comparative evaluation of ML/DL algorithms (Scikit-learn, TensorFlow, PyTorch).◦ Creation of Python APIs to expose recommendation functionalities on the Disneyland Paris website.◦ Cloud deployment: Orchestration of the system with Docker on AWS (SageMaker, ECS, S3, EC2).◦ Document ingestion for RAG: Implementation of an extraction system via AWS Glue to process technical documentation for Disney's entire application portfolio.◦ Vectorization: Development of the semantic vectorization layer with SageMaker and LlamaIndex, and indexing with OpenSearch for precise searches.◦ Configuration of the RAG pipeline with Claude 3.5 Sonnet on Amazon Bedrock, including prompt engineering and contextual management.◦ Writing technical documentation.
Recommendations
Be the first to recommend Tarik
Help this freelancer shine by sharing your experience working together.
These freelancer profiles also match your criteria
Agatha Frydrych
Backend Java Software Engineer
4.7
(3)
2
Baptiste Duhen
Fullstack developer
4.6
(4)
5
Amed Hamou
Senior Lead Developer
4
(2)
7
Audrey Champion
Web developer
4.3
(3)
4
Education
- Machine Learning SpecialisationCoursera2022Machine Learning Specialisation
- State Engineer Diploma in Operational Research and Decision SupportNational Institute of Statistics and Applied Economics2016Diplôme d'Ingénieur d'Etat en Recherche Opérationnelle et Aide à la Décision