[전문연구요원] Machine Learning Engineer – LLM/ RAG

Roles and Responsibilities 

• RAG Model Research & Development 
• Research, design, and implement cutting-edge Retrieval-Augmented Generation algorithms and methodologies. 
• Continuously evaluate and improve model effectiveness and accuracy. 
• Large Language Models (LLM) Optimization 
• Fine-tune and optimize state-of-the-art LLMs to enhance retrieval performance and relevance. 
• Develop methods to reduce latency and improve model responsiveness in production environments. 
• Semantic Search & Vector Database Integration 
• Develop and enhance semantic search capabilities, utilizing our novel vector databases solution 
• Optimize retrieval mechanisms for maximum performance and scalability. 
• Data Management & Pipeline Development 
• Create and manage data pipelines for training, evaluating, and continuously updating RAG models. 
• Ensure high-quality training datasets through advanced data preprocessing and cleaning techniques. 
• SaaS Application Integration 
• Lead the integration of RAG and LLM models into production SaaS solutions. 
• Collaborate closely with software engineering teams to ensure seamless deployment and scalability. 

Basic Qualifications 

• Bachelor's or Master's degree in Computer, Machine Learning, or a related field. 
• Extensive hands-on experience with Large Language Models. 
• Deep expertise in Retrieval-Augmented Generation (RAG) and semantic search. 
• Proficiency in Python and familiarity with ML frameworks (PyTorch, TensorFlow, Hugging Face). 
• Strong foundation in data structures, algorithms, and distributed computing. 
• Effective collaboration, communication, and problem-solving skills 

Preferred Qualifications 

• Advanced degree (PhD preferred) in Computer Science, Machine Learning, or a related field. 
• Contributions to research publications or open-source communities focused on NLP, RAG, or related fields. 
• Experience optimizing models for deployment on cloud platforms (AWS, GCP, Azure). 
• Familiarity with MLOps practices, CI/CD pipelines, and model monitoring.