Freelancer profile translated to English.
Back to original languageAbout Erwan
Hello,
Software engineer specializing in generative AI, I design SaaS applications and systems integrating generative AI, with particular expertise in RAG (Retrieval-Augmented Generation) architectures and LLMs in production.
I am involved in the entire chain: product design, backend and frontend development, data structuring, setting up AI pipelines (embeddings, semantic search, agents), up to the deployment of robust and scalable solutions.
I have developed several projects combining SaaS and AI:
Ada(Ministry of Culture X Cité des Sciences): platform for indexing and searching multimodal educational content.
Document AI(Ministry of Ecological Transition): evaluation of structured information extraction pipelines.
U31.io: intelligent text simplification editor based on LLMs.
I also provide training on generative AI and RAG topics, supporting technical teams in setting up complete pipelines (from data ingestion to integrating LLMs in production).
->I enjoy understanding product challenges and proposing concrete, effective, and adapted solutions, rather than simply executing.
I would be delighted to discuss your project with you.
French
Native or bilingual
English
Fluent
Remote only
Primarily works remotely
Experience
- Ministère de la CultureFull Stack ML DeveloperPUBLIC SECTORJune 2024 - June 2025 (1 year)Paris, FranceAda aims to provide teachers with Universcience's (Cité des sciences / Palais de la découverte) scientific multimedia heritage, enhanced by AI, to transform science education.My role as the sole developer on the project:- Semantic indexing: PostgreSQL + pgvector: creation of the multimodal data analysis pipeline- AI Analysis: Whisper (video/audio), Florence-2 (images), open source LLMs- Web: Next.js, Chatbot UI, course editor via Tiptap- Backend: Python + FastAPI
- U31.ioCTOSOFTWARE PUBLISHINGOctober 2019 - April 2024 (4 years and 6 months)Paris, FranceOne in six people have difficulty understanding your content (INSEE).U31 functions like a spell checker, but for text simplicity: it analyzes and reformulates complex passages to make them more accessible.Before the advent of modern LLMs, simplification relied on classic linguistic rules (Python + SpaCy).With the arrival of GPT-3.5 (late 2022), initial tests showed potential, but the quality remained insufficient for our needs.We then conducted in-depth R&D: selecting open-source LLMs, fine-tuning on a proprietary dataset, and achieving simplification performance superior to proprietary models.Linguistic analysis: Python + SpaCyApplications: React.JS + Tiptap for interactive editingDatabase: MongoDBAI: LLM Fine-tuning to simplify and rephrase text
- Seed-UpFull Stack DeveloperTECHOctober 2016 - November 2019 (3 years and 1 month)Paris, FranceDevelopment of several projects:Magic Lamp: design of a lamp for parents (mini-projector, raspberry) that generates, animates, and projects stories told in real-time using voice recognition and natural language analysis.Fiters: creation of a matching API between coaches and athletes (Uber-type API, Node.JS, PostgreSQL, Stripe, Google APIs)kronenbourg: Prototyping a connected beer dispenser (Arduino, DIY, NFC)
Reviews
Recommendations
Be the first to recommend Erwan
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
- Master, Software Architect Application DeveloperETNA2019Master en informatique après 5 années à l'ETNA (dont deux de prépa). L'ETNA s'appuie sur la pédagogie par projet (type hackathon) 3,5 ans en entreprise.