Objectives

• Deploy and benchmark open-source LLMs locally to evaluate performance, latency, and resource usage.
• Fine-tune models on custom and confidential datasets while preserving data privacy and confidentiality.
• Integrate Retrieval-Augmented Generation (RAG) to improve context-based reasoning and answer accuracy.
• Build an intelligent AI agent capable of secure query handling and autonomous task execution.
• Ensure a privacy-preserving, offline-capable architecture suitable for sensitive environments.

Required Skills

• Deep understanding of Large Language Models (LLMs) such as GPT, BERT, LLaMA and their architectures.
• Experience in fine-tuning, deploying, and optimizing AI models for inference and resource constraints.
• Strong programming skills in Python and proficiency with the Transformers library and PyTorch.
• Familiarity with RAG pipelines, vector databases, and database management systems such as PostgreSQL.
• Knowledge of AI agents, autonomous system design, and secure query handling practices.
• Curiosity, adaptability, and commitment to stay up to date with the latest AI advancements.

Tasks & Deliverables

• Set up local deployment pipelines for multiple open-source LLMs; run systematic benchmarks (throughput, latency, memory, and accuracy).
• Design and execute fine-tuning experiments on custom/confidential datasets with attention to data protection and reproducibility.
• Implement RAG workflows: document ingestion, vectorization, retrieval, and integration with language models; evaluate impact on reasoning.
• Build and validate an AI agent that can handle queries securely (access control, query sanitization) and operate offline when needed.
• Integrate and manage vector database(s) and PostgreSQL for persistent storage, retrieval performance tuning, and backup strategies.
• Produce reports, reproducible code, model cards, and documentation for deployment, benchmarking results, and privacy measures.

Application

• To apply, send your CV and a brief cover note describing relevant experience (fine-tuning, RAG, deployments) to hr@iovision.io .
• Highlight past projects or repositories demonstrating LLM work, PyTorch/Transformers usage, or RAG/vector DB integrations.
• Use the email subject: "Application for 05 07 06 Objectives PFE" when contacting the recruiter.