// Freelance · Adone LTD

I build ML models, and ship them to production.

Data Scientist & DevOps Engineer. From model to production: RAG/LLM, observability, CI/CD and scaling.

machine learning · mlops · devops

Available for freelance work · Remote / France · Day rate on request

Get in touch → View projects Download CV ↓
network · training → deployment

About

Between data science and operations.

Freelance Data Scientist and DevOps Engineer with around five years of experience taking machine learning from notebook to production. I design models — forecasting, NLP, LLM/RAG, computer vision — and build the pipelines, containers and monitoring that keep them running reliably.

My focus: industrialising models, integrating RAG/LLM systems, and keeping the observability and CI/CD around them solid. Trilingual, I work in French, English and German.

5+ yrsExperience
MScData Science & ML Eng.
AI-900 · AI-901Microsoft Certified
FR · EN · DETrilingual

Stack

Model, generate, code, deploy.

Machine Learning & Data Science

Languages & frameworks

  • Python
  • PyTorch
  • TensorFlow
  • scikit-learn

LLM & RAG

  • Mistral
  • LLaMA 2
  • BERT
  • Pinecone

Modelling

  • Time-series forecasting
  • Regression
  • Clustering
  • Vision · OpenCV

Generative AI

LLM & RAG

  • RAG
  • Embeddings
  • Vector search
  • Pinecone

Agents & orchestration

  • LangChain
  • Agents
  • Function calling

Adaptation

  • Prompt engineering
  • Fine-tuning
  • Evaluation

Python Engineering

Code & architecture

  • Clean code
  • Clean architecture
  • SOLID
  • Type hints

Testing

  • pytest
  • TDD
  • Coverage
  • Mock

Backend

  • FastAPI
  • Flask
  • REST API
  • Pydantic
  • async

MLOps & DevOps

Containers & orchestration

  • Kubernetes
  • Docker

Integration & deployment

  • CI/CD
  • Azure DevOps/TFS
  • pytest

Observability & streaming

  • Prometheus
  • Grafana
  • Thanos
  • Kafka
  • Flask API

Data Engineering

Reliable data, at scale.

I build the pipelines that feed ML: real-time ingestion, transformation, storage and serving — orchestrated, tested and monitored.

Sources

Data from everywhere: APIs, SQL/NoSQL databases, application logs and real-time streams.

  • REST
  • SQL
  • Logs
  • Streams

Data Visualization

Making data speak.

Clear dashboards and visualisations that turn raw numbers into decisions.

Sample data · 12 months

Experience

From model to scale.

2024 — Present

Data Scientist & MLOps · Adone LTD (Freelance)

Designing and deploying ML and LLM systems for a range of clients:

  • Luxury group — RAG/LLM service integrated into Kubernetes microservices.
  • Regional press group — audience analytics and NLP on editorial content.
  • Metropolitan transport network — demand forecasting and anomaly detection on real-time data.

2021 — 2024

Data Scientist / DevOps · Peaksys (Cdiscount)

Observability-driven ML: log anomaly detection (OpenSearch), time-series forecasting, model industrialisation and CI/CD over Kubernetes, Prometheus, Grafana, Thanos and Kafka.

2021

Data Analyst — Internship · Satelia

Data analysis and reporting on remote patient-monitoring topics.

Education

Degrees, certifications & languages.

2019 — 2024

Master 2 — Data Science & ML Engineering

Ynov Informatique · Bordeaux

Computer-engineering programme, specialised in data science and the industrialisation of machine-learning models.

2018

Baccalauréat — Literature & Mathematics

Lycée Labourdonnais · Curepipe, Mauritius

Demo

Semantic search & RAG.

Ask a question: the query is vectorised, then compared against a knowledge base. The closest passages surface with their similarity score — the heart of a RAG system. Runs 100% in your browser.

Try:

    Demo

    Computer vision, live.

    A vision pipeline running right in your browser, no server: grayscale, edge detection (Sobel) and detection boxes — in real time.

    Runs 100% locally — no camera, no server.

    Methods

    Agile, end to end.

    I work Scrumban: the cadence of Scrum (sprints and ceremonies) combined with the continuous flow of Kanban, wired to an MLOps CI/CD — from backlog to production monitoring.

    Sprint planning

    Plan

    The backlog is sliced into estimated tickets (Scrum) and prioritised as a flow (Kanban); we commit to the iteration.

    • Azure Boards
    • Jira
    • User stories

    To do

    In progress

    Done

    Tip: click a ticket to move it along the flow.

    Observability

    Monitoring, live.

    An interactive peek at what I run in production: real-time metrics, anomaly detection and a log stream — like a Grafana dashboard.

    adone-ltd · prod

    Projects

    Selected work.

    LLM · RAG

    RAG platform on Kubernetes Confidential

    Retrieval-augmented LLM service embedded in microservices for a luxury group. Vector search, orchestration and API exposure.

    • LLM
    • RAG
    • Pinecone
    • Kubernetes
    • Flask

    Observability · ML

    Anomaly detection in production

    Anomaly detection and forecasting over production logs, surfaced in Grafana dashboards for operations teams.

    • OpenSearch
    • Prometheus
    • Grafana
    • Kafka

    Computer Vision

    Restoring an old photograph

    Restoring a family photograph with classical image processing: denoising, correction and reconstruction of damaged areas.

    • OpenCV
    • Python
    • Image processing

    Identity · Design

    Adone LTD brand system

    A full visual identity built around a neural-network symbol and an MLOps motif (CI/CD loop): logo, signature, business card.

    • SVG
    • Design
    • Branding

    Contact

    Let's build something.

    Available for freelance machine-learning and MLOps work.