Project overview

• Implement an AI-based Knowledge Capitalization System to centralize and organize the project team's expertise (technical documentation, meeting reports, etc.).
• The system must allow searching, summarizing and extracting relevant information from heterogeneous technical sources to facilitate reuse of know-how.

Objectives & main missions

• Design and develop the platform architecture (ingestion, indexing, retrieval, summarization and UI) and deliver a working prototype.
• Implement pipelines to ingest documents and meeting reports, extract embeddings, index them (Elasticsearch or PostgreSQL + pgvector) and expose search/summarization APIs.

Technical stack & tools (as specified)

• Backend: Node JS (Express); IA / NLP components: Python (LangChain JS); Vector search: Elasticsearch or PostgreSQL + pgvector.
• Frontend: React; Object storage: MinIO or AWS S3; other tools: CI/CD, containerization and testing as needed.

Responsibilities & expected deliverables

• Deliver a searchable indexed corpus, REST APIs for retrieval and summarization, a React-based UI for end-users, and technical documentation.
• Provide evaluation results (precision/recall or relevance metrics), examples of extracted summaries, and deployment instructions.

Required profile

• Engineer profile with skills in NLP/AI, Python, experience with LangChain-style toolkits, and familiarity with vector databases (Elasticsearch or pgvector).
• Web/fullstack skills: Node.js (Express) and React; experience with object storage (MinIO or AWS S3); good software engineering practices (tests, CI/CD).

Project specifics & logistics

• Reference: PRX-2026-13 (Proxym Group).
• Number of positions: 2 trainees; expected duration: 3-6 months (PFE), with a focus on producing a demonstrable, deployable prototype.

How to apply

• Apply online via the trainees platform: https://trainees-platform.proxym-group.net
• In your application, mention the reference PRX-2026-13 and highlight relevant projects in NLP, vector search and fullstack development.