RAG Algorithm Engineer

Responsibilities Optimize the entire pipeline of the Retrieval-Augmented Generation(RAG) system in the financial domain, including multi-source heterogeneous data processing (e.g., financial reports, market sentiment, and market data), knowledge semantic modeling, hybrid retrieval (semantic/keyword/structured retrieval), reasoning-based retrieval, Rerank model optimization, and enhancing large model generation performance; Design and implement RAG solutions for use cases such as intelligent investment advisory and risk management, tackling key challenges like multimodal fusion, multi-hop reasoning, noise filtering, fine-grained real-time traceability, and more; Build a high-precision financial retrieval system by integrating knowledge graphs and vector embedding technologies while continuously optimizing precision-recall, response efficiency, and dynamic data adaptability; Research cutting-edge RAG technologies (e.g., real-time knowledge updates, fine-tuning large models) to drive innovative applications in financial scenarios; Enhance the cooperative mechanisms between large language models (LLMs) and vector models to improve financial text processing capabilities; Establish a closed-loop system for the validation of technological optimizations and business value, and design performance evaluation frameworks. Position Requirements Master's degree or higher in Computer Science, Artificial Intelligence, or related fields, with at least 3 years of experience in optimizing and architecting large-scale RAG systems; Proficient in LLM fine-tuning techniques (e.g., LoRA, P-tuning), prompt engineering, and generation control; familiar with retrieval models such as BERT/ColBERT and vector stores like FAISS/Elasticsearch, as well as optimization of LangChain/LlamaIndex frameworks. Experience with optimization on datasets containing billions of vectors or trillions of tokens is a plus; Thorough understanding of financial text characteristics (e.g., financial reports, research reports) and familiarity with Wind Financial Terminal is a plus; Proficient in Python/Java/Scala and capable of developing high-concurrency retrieval systems; Demonstrated innovation in areas such as query expansion, semantic ranking, and multimodal fusion; a record of publications in top-tier academic conferences or ownership of related patents is an advantage. Preferred Qualifications Experience with building financial knowledge graphs, with expertise in entity and relationship extraction and reasoning; Participation in large-scale algorithmic projects like intelligent investment advisory or risk management engines; Familiarity with distributed GPU training and computational efficiency optimization. Details Seniority level: Mid-Senior level Employment type: Full-time Job function: Engineering and Information Technology #J-18808-Ljbffr

Created: 2025-09-21