Algorithm Engineer - Large Model

Job Responsibilities:

• Participate in the company's decision-making large model product continuous optimization design and implementation, decision-making large model training data construction, data matching, model training and indicator evaluation and other related work.
• Responsible for the training and research of large models, including but not limited to pre-training, SFT, Alignment, multimodal technology, etc., and explore the application of large models in the semiconductor field.
• Build a proprietary large model in the semiconductor field, integrate semiconductor field knowledge, and quickly implement semiconductor business scenarios.
• Responsible for researching and understanding the large model test system and data construction methods, and evaluating the full capabilities of large models in understanding, reasoning, agent, rag, etc.
• Responsible for the overall design and development of the LLM model reasoning engine, optimize the engine, algorithm and model architecture, and improve computing performance.
• Responsible for the exploration of cutting-edge natural language understanding technologies, including model distillation, edge computing, multi-round semantic understanding, knowledge fusion, etc.
• Responsible for following up on the latest progress in the field of large models, understanding the cutting-edge trends in the industry, and continuously iterating large model-related algorithm modules to achieve product vision and goals.

Job Requirements:

• Computer science, artificial intelligence and other related majors, master's degree or above, doctoral degree preferred. Have enthusiasm and confidence in the future development of large model technology.
• Master the python coding language, and be proficient in one or several deep learning frameworks (such as tensorflow, pytorch, paddlepaddle, etc.).
• Master the basic theories and algorithms of NLP, have relevant industry experience in NLP, and be proficient in the architecture and principles of the Transformer series of models.
• Master natural language processing tasks, such as question-answering systems, retrieval systems, recommendation systems, knowledge graphs, reasoning graphs, sequence annotation, summary generation and extraction, etc.
• Have a deep understanding of the mainstream pre-trained language large models of NLP (GPT/LLaMA/GLM/Bloom/Qwen/BERT), and have strong exploration and practical skills.
• Master the relevant technologies of large models, and have a certain understanding and exploration experience of mainstream large model application frameworks such as RAG, AI-Agent, LangChain, etc.
• Master the training data construction, data ratio, training and indicator evaluation of large models PT, Continue PT, SFT, and RLHF in vertical industry fields.
• Familiar with at least one large model distributed training framework such as DeepSpeed, Megatron, Colossal-ai, etc.
• Familiar with LLM-related inference engines and their mainstream optimization methods, such as /Triton/FasterTransformer/vLLM, FlashAttention/PageAttention, etc.
• Excellent English reading and writing skills, fluent listening and speaking skills.
• Those with overseas study background are preferred.

Send your CV to:  hr@alphaxtec.com