Ancheng Wu
CTO of Diffractive Number Technology,Author of Neural Networks and Deep Learning
Ancheng Wu , CTO of Beijing Diffractive Digital Technology, a technical expert with more than 20 years of R&D experience, has authored "Neural Network and Deep Learning", "Deep Learning Algorithms in Practice", "The Secret of Trading - Making the First Bucket of Money with Algorithms" and other books. He has deep technical accumulation and wide technical stack in the fields of artificial intelligence, big data and deep learning; he has different application system landing experience in the algorithmic direction of text, speech, recommendation and image; he owns 11 patents in algorithmic category; he has won the first place in reading comprehension competition Squad2 single model; in the field of big model, he has led a series of important projects, including ChatGPT-like big language base model training and optimization, as well as applications in finance, insurance, cultural creation and Chinese medicine, etc. He has led the team to complete the training of language models from 1.3B to 13B, and carried out the fine-tuning of business data from 30B to 63B large models.
Topic
Vertical Industry Large Model Engineering Practices
This presentation will cover the following important topics: Big Model Data and Training: we will discuss how to build big models based on enterprise vertical data, how to choose appropriate training methods, and share strategies on how to effectively evaluate model performance. Big Model + RAG: We will discuss how to leverage hierarchical data and how to choose the right RAG model to enhance the effectiveness of big models. Also, we will dive into the two key aspects of recall and ranking. Agents and Beyond: In this section, we will compare different big models and discuss their alternatives, as well as share how to effectively apply multiple Agents in real-world scenarios. We look forward to working with you to delve into these topics and share our experience and insights from our engineering practice in landing big models in vertical industries.
