Machine Learning
Understanding Isolation Forest for Anomaly Detection
Isolation Forest is one of those machine learning algorithms that seems almost too simple, yet surprisingly powerful.
Oct 7, 2025
01
About Me
Passionate about AI and data science for several years, I've specialized in machine learning, deep learning, and more recently Generative AI and Agentic systems.
My journey has led me to design complex ML pipelines, build production-grade models, and deploy AI solutions on cloud infrastructures. I combine strong software engineering foundations with cutting-edge AI research.
Today, I develop end-to-end AI systems including recommendation engines, NLP solutions, computer vision applications, and intelligent agents (RAG, Multi-Agent Systems).
Tech Stack
Python / PyTorchTensorFlow / KerasLangChain / LangGraphScikit-learn / XGBoostDocker / KubernetesAzure / GCP
02
Work Experience
AI Developer
Freelance
November 2025 - Present
- —Design of custom solutions combining software engineering and artificial intelligence for private and independent clients
- —Development of data-driven web applications and AI-based automation tools (workflows, AI agents, NLP, Langchain, Langgraph)
- —Integration of AI models into existing applications (APIs, ML pipelines)
- —Deployment and production of robust and scalable solutions (Docker, CI/CD, cloud)
PythonNextJSFastAPILangChainDockerSupabase
Data Scientist (Internship)
SNCF Réseau
March 2025 - September 2025
- —Assisting business teams in data scoping and AI use case definition
- —Design and industrialization of Big Data pipelines for anomaly detection on 500M+ time-series data points
- —Selection and implementation of Deep Learning and Machine Learning algorithms (LSTM, Isolation Forest) with functional and technical analyses
- —Optimization of data and algorithmic processing: -40% reduction in computation time and improved stream reliability
- —Monitoring models in production and tracking performance over time
PythonTensorFlowScikit-learnAzure DatabricksPySparkMLflow
R&D Data Engineer (Internship)
EffetB
April 2024 - August 2024
- —Development of a hybrid recommendation system, combining knowledge graphs (Neo4j) and advanced NLP
- —Leveraging community detection to better segment users and strengthen recommendation quality
- —30% reduction in cold start issues through integration of contextual embeddings (TF-IDF, Keras embeddings)
- —Implementation of advanced algorithms: GraphSAGE, FastRP, KMeans, t-SNE dimensionality reduction with 3D visualizations
- —POC validation allowing product integration and concrete user-oriented use cases
Neo4jPythonKerasGraphSageFastRPScikit-learn
Data Scientist
Intside LLC
December 2021 - August 2023
- —Analysis of user behavior, web traffic and marketing campaign performance to drive decision-making
- —Design of a clustering model (KMeans) that identified 4 strategic client segments
- —Design of lead scoring models integrated into the agency's targeting strategies
- —Creation of Power BI dashboards and cross-functional collaboration with product, tech and marketing teams
PythonScikit-learnPower BISQLGCPKMeans
Software Developer / Data Analyst
Alibora SARL
December 2021 - August 2022
- —Built a call-center agent management system with performance tracking and audio integration
- —Enhanced a property-management mobile app and contributed to ETL pipelines
- —Developed REST APIs using Laravel for data processing workflows
DockerJavaScriptLaravelSQLAPI Development
03
Skills & Technologies
Languages
PythonSQLRTypeScript
3+
Years Experience
15+
Projects Done
20+
Technologies
04
Featured Projects
Kontact AI
A Saas platform that help to prospect on Linkedin
Agent IAOpenRouteur AIAutomatisationSupabasePythonNextJS
Emotional Report
AI system that detects your emotion based on your voice recordings. Uses Wav2Vec2 for audio feature extraction and LSTM networks for emotion classification.
PythonWav2Vec2LSTMStreamlitDeep Learning
More Projects
Vendora
E-commerce platform with recommendation engine using collaborative filtering.
Next.jsTypeScriptAPI REST
Anomaly Detection Pipeline
LSTM Autoencoder + Isolation Forest for time-series anomaly detection on 500M+ records.
PythonTensorFlowPySparkAzure
Knowledge Graph Recommender
Hybrid recommendation system using Neo4j knowledge graphs and NLP embeddings.
Neo4jGraphSagePythonKeras
RAG Document Assistant
Retrieval-Augmented Generation system for intelligent document Q&A.
LangChainOpenAIVector DBFastAPI
06
Let's Connect
I'm always open to new opportunities, collaborations, or just a chat about AI, data science, and tech. Feel free to reach out!
Get in Touch