Freelancer profile translated to English.
Back to original languageAbout Sébastien
Data Scientist / Data Engineer / FullStack Web Developer
Paris / Grand-Est | 25 years of experience |
Experience with large companies and startups
LinkedIn Profile : smougel
Artificial Intelligence & Machine Learning
Generative AI & LLM
Autonomous AI Agents (Tool Calling, Human-in-the-loop)
Custom Conversational Agents (RAG, Vector Databases, LangChain)
Multi-provider inference and sandboxing (Docker)
NLP & Deep Learning
Transformer Architecture (BERT, GPT)
NER, embeddings, semantic search
Image Analysis (CNN, ResNet)
Machine Learning
Industrial-scale modeling and training (AWS, Azure)
Classification, clustering, anomaly detection
Recommendation Systems
Development & Infrastructure
Python Development and APIs
High-performance Fullstack Web (High Traffic Websites)
Asynchronous processing, batch, and parallelization
Data Collection and Pipelines
Project Approach
Scoping workshops and needs analysis
Data audit and exploration
Co-construction of the technical solution
Team training
English
Fluent
French
Native or bilingual
Remote only
Primarily works remotely
Experience
- ArdianExpert NLPBANKING AND INSURANCESeptember 2024 - Today (1 year and 9 months)Paris, France
Automation of contract analysis by AI
Context: Industrialization of the extraction and classification of legal clausesSolution: Design of an end-to-end LLM system with anti-hallucination mechanisms and parallelized inferenceResults: Production pipeline validated by benchmark (Precision/Recall/F1-Score metrics)ESG Barometer
Context: Automated analysis of ESG commitments from corporate communicationsSolution: Automatic extraction and categorization of ESG themes from emails, financial documents, and attachments via LLMImpact: Real-time monitoring of ESG trends for decision supportFinancial Intelligence Platform
Context: Large-scale automated processing of financial documentsTechnologies:Text analysis models (PyTorch: LSTM, GRU)Document classification (Doc2Vec)Trend and emerging topic detection (BERTopic, LDA)Impact: Acceleration of processing and extraction of actionable insightsSignature and Structured Data Extraction Solution
Context: Automation of professional email processingSolution:Automatic detection of email signaturesNamed Entity Recognition (NER with Flair, LLM)Impact: Automatic structuring of unstructured data for CRM enrichment - Beyond1CEO of Beyond1TECHMay 2022 - Today (4 years and 1 month)
Design and development of HIGGINS
HIGGINS - AI Agent Orchestration PlatformClient-server architecture • Python/Flask • JavaScript • Redis • DockerConceptIntelligent orchestration solution: automatic delegation of tasks to the most relevant specialized AI agentsTechnical skills utilizedBackend: Python, Flask, Microservices Architecture, Dependency InjectionAI/LLM: Multi-agent orchestration, MCP Protocol, Prompt engineeringReal-time: Redis Pub/Sub, Server-Sent Events, WebSocket, StreamingFrontend: JavaScript, Conversational Interface, Minimalist UXDevOps: Docker, Secure Sandbox, Isolated environment managementKey achievementsIntelligent meta-agent: routing system analyzing requests and selecting the optimal agent from a library of specialized agentsNative observability: real-time streaming with full visibility of agent actions, decision traceability, displayed confidence scoresMCP Protocol: integration of the emerging standard for connecting to external servers and dynamically extending capabilitiesHuman validation: pre-execution control system for each critical action, ensuring safety and controlProduction-ready architecture: graceful degradation mechanisms, automatic transfer to human supervision, transparent error handlingSEOCopilot Editor
https://www.seocopilot.frSaaS software that helps improve SEOProvides decision support for content writing, semantics.Tools based on data and machine learning.Keyword opportunity research through semantic search, cross-referencing data from Google Search Console.Conversion rate prediction.Use of LLMs (Large Language Models): CamenBERT, Sentence Bert, Chat GPT Fine Tuning, T5Word & Sentence Embeddings: FastText, Word2Vec, SentenceBert - ENGIE - SolutionsData EngineerENERGY AND UTILITIESDecember 2021 - April 2022 (5 months)Rueil-Malmaison, FranceEngie SAS Data Platform is in charge of collecting data from millions of IoT sensors installed in France.As a consultant for Engie, my mission was to work on improving the scalability and stability of the data platform into Azure ecosystem.My realizations :- Analysis of Application Behavior (Bottleneck identification)- Improvements of Azure SQL queries through analysis of query execution plan (SQL Server)- Improvements on Azure functions efficiency- Development of an app component in charge to improve the processing of database queries.- Building tools for supervision and monitoring : Azure Dashboard, Alerts & Python scripts to reports KPIs.- Improved software architecture (Data Buffer)- Cost optimization of azure resources : Application Insight / Azure Storage.
Reviews
Recommendations
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
- Software Engineering EngineerCNAM2008
- AI Engineer - Data ScienceOpenclassrooms - Microsoft2021Analyse de données, Modélisation, Mise en production de modèles dans le cloud (Microsoft Azure), Statistiques, Utilisation des principaux frameworks (TensorFlow, Keras, Scikit-learn)
Certifications
- Machine learningCoursera2020
- Deep learning specializationCoursera2020