Project overview

• Develop an intelligent RAG-based module for a banking platform that automatically generates official documents and client summaries.
• The module must guarantee accuracy, traceability and compliance while integrating human validation steps into the pipeline.

Objectives & impact

• Reduce document processing time and increase productivity by automating generation and summarization tasks using Retrieval-Augmented Generation (RAG) and LLMs.
• Ensure traceability and regulatory compliance by maintaining audit trails and explicit human approval points for produced documents.

Responsibilities / Scope of work

• Design and implement the RAG + LLM pipeline (retrieval, context assembly, prompt handling, generation, post-processing) and integrate it with the banking platform.
• Build APIs and microservices to support document generation, summarization, versioning, human validation workflow and logging for traceability.

Expected deliverables

• A working prototype demonstrating secure document generation and client summary workflows with human validation and audit logs.
• Documentation, tests, and deployment artifacts showing how the module integrates with existing systems and storage (PostgreSQL, MinIO).

Required profile & conditions

• Engineer profile / Bachelor’s degree. Required profile: 1 Trainee.
• Good knowledge/interest in NLP, RAG, LLMs, prompt design, and secure software development for regulated environments.

Technologies & tools

• Frontend: React JS (integration components for validation/review flows).
• Backend & APIs: NestJS, Python microservice for ML tasks; database and object storage: PostgreSQL, MinIO.
• Other expected tools: containerization (Docker), CI/CD, logging/monitoring, and basic DevOps practices to deploy microservices.

Security, compliance & integration

• Implement data privacy and secure handling of client data, maintain audit trails for every generated/sent document.
• Ensure seamless integration with existing banking systems and strict human-in-the-loop checks prior to finalizing official documents.

How to apply

• Apply via the trainees platform: https://trainees-platform.proxym-group.net
• When contacting recruiters, reference REF: PRX-2026-09 and the project title in your message.