Freelancer profile translated to English.

Description

Data Scientist and AI Engineer with over 6 years of experience in designing, industrializing, and deploying data & AI solutions with high business value. Specializing in Generative AI (LLMs, RAG, agents), I design scalable, secure, and production-oriented platforms, from the scoping phase to cloud deployment.

I recently led the development of a sovereign AI platform for a major insurance group, integrating advanced RAG architectures, orchestrated LLM agents (LangGraph), and APIs exposed via secure services. My expertise covers the entire chain: data engineering, machine learning, GenAI, MLOps, and DevOps, with strong command of GCP and AWS environments.

Languages

Arabic
Native or bilingual
French
Fluent
English
Fluent

Workplace preferences

Can work on-site

Paris (up to 50km)

SFEIR
Data Scientist AI / GenAI
February 2024 - Today (2 years and 4 months)
Paris, France
Design and development of a sovereign AI platform ensuring the confidentiality and control of internal data:

- Design of a RAG system based on LangChain, capable of extracting and indexing heterogeneous documents (Excel, CSV, PDF, etc.).
- Integration of LLMs via Vertex AI, ensuring the generation of contextualized responses and customization according to the group's businesses.
- Cloud-native orchestration on GCP, combining Cloud Run for scalable inference, Cloud SQL (PostgreSQL) for embedding management, and Vertex AI for model usage. – MLOps & DevOps Industrialization: unified CI/CD pipelines with Terraform (IaC) and Jenkins, ensuring reproducibility and continuous delivery of environments.
- Centralized monitoring via BigQuery and Looker Studio, allowing real-time tracking of logs, API calls, and error rates.
- Secure integration into AG2R's internal ecosystem: APIs exposed on a private network and consumed by the ALMIA BOT front-end, adopted by over 6000 employees.
- Strategic opening of AI services via Apigee, facilitating their integration into the group's business applications.

Leadership of the second phase of the project, dedicated to integrating an agentic AI architecture aimed at strengthening the system's autonomy and orchestration capabilities:

- Design and implementation of a central orchestrator agent, responsible for coordinating the execution of several specialized agents developed with LangGraph.
- Deployment and exposure of agents on a LangGraph server hosted in Cloud Run, ensuring automatic scalability.
- Monitoring and evaluation of agent performance, including coordination latency, exchange robustness, and relevance of autonomous decisions.
Cloud Engineer AI Agent Generative AI Data Science CI/CD
Heuritech
Data Scientist
December 2021 - January 2024 (2 years and 1 month)
Paris, France
Analysis of image data to anticipate fashion trend evolutions:

- Define a new normalization approach to improve trend representativeness.
- Develop a mini-dashboard to allow business units to monitor year-over-year (YoY) growth.
- Study the relevance of trends (noisy time series where the signal-to-noise ratio is too low).
- Implementation of a new evaluation module based on scale-free-error metrics.
- Define a new holistic approach to correct the bias of the individualistic forecast model.
- Replace Python scripts executed locally with Prefect flows to ensure efficient and reliable workflow management.

Migration of data from Elasticsearch to the Snowflake data warehouse to improve trend collection scalability:

- Guarantee data integrity throughout the migration process.
- Implement trend query logic in SQL for a smooth transition to Snowflake.
- Significant optimization of query performance, implementing improvements to ensure more efficient execution.
- Modification of the trend pipeline's Prefect flow to direct the flow to Snowflake, thus consolidating performance improvements.
Python Time Series Forecasting Data Migration Data Analysis
Datasprints
Data Scientist / Data Engineer
January 2020 - November 2021 (1 year and 10 months)
Belo Horizonte, MG, Brazil
Development of a recommendation system (user preferences service) for the medical tele-expertise platform:

- Develop a data migration pipeline (OLAP - OLTP): regular copying from Redshift to Aurora.
- Prepare training and testing datasets: remove missing observations, encode features, etc.
- Design and implementation of a recommendation engine based on Word Embedding techniques.

Development of a classification model to distinguish repetitive complaints from new ones:

- Collect data from various sources (on-premise, database, data warehouse), design a Common Data Model (CDM), transform extracted data to match defined standard entities.
- Conduct an exhaustive exploratory study and share key insights with business experts via a dashboard.
- Extract handcrafted features and use open data to enrich the dataset.
- Hyperparameter tuning including "cost-sensitive learning" parameters to manage data imbalance.

Review of tire spending performance & provide savings measures for PRAXIO:

- Create and maintain Business Intelligence (BI) dashboards.
- Lead a team of analysts to integrate a data-driven culture within the company.
- Conduct in-depth analytical understanding and provide actionable insights for better tire utilization.

Optimization and deployment of a statistical model developed by the R&D team to fine-tune and centralize the validation process for industrial pipes:

- Accelerate Monte Carlo simulations using Numba and Dask.
- Design of a scalable and serverless infrastructure hosted in the cloud to contain the model and automate its deployment, as well as the deployment of the model in AWS using Terraform.
Cloud Engineer Data Science Data Migration Big Data