Kai Zhang
Deputy Director of the Financial Intelligence Special Committee, Shanghai Open Source Information Technology Association
Ph.D. in Artificial Intelligence from Xi'an Jiaotong University, with over 15 years of research experience in AI and big data. He has been deeply involved in algorithm development and team management in the finance and industrial manufacturing sectors, with in-depth research and practical experience in areas such as large model pretraining and application, personalized internet recommendation, and ad-serving algorithms. He has repeatedly built data and algorithm teams from scratch. He leads teams in building AI platforms and has established dozens of standardized capabilities in text parsing, image recognition, and speech recognition. By integrating traditional AI with large language models, he has driven the implementation of AI products and intelligent services across a wide range of scenarios, including financial data governance, intelligent customer service, AI-powered investment research and advisory, sentiment analysis, risk control and anti-fraud, personalized recommendation, and intelligent search. His work has enabled enterprises to reduce costs and increase efficiency, with proven capability in designing and delivering products that serve tens of millions of daily active users.
Topic
Open Source Empowerment: The Development and Practice of Trustworthy Financial Agents
This talk focuses on the core theme of "Open Source Empowering Trustworthy Financial Agents." With the deep integration of artificial intelligence into the financial sector, financial agents—intelligent entities that combine perception, reasoning, and planning—have emerged as a key form of application. Their “trustworthiness” (security, compliance, explainability, etc.) has become a critical prerequisite for industry adoption. The talk will explore how open collaboration, transparent mechanisms, and community efforts can support the technological development, standard formulation, and practical deployment of trustworthy financial agents. At the same time, it will analyze the specific value of enhancing financial agents' information traceability and system reliability through real-world practices such as RAG technology and modular architectures, while objectively examining the challenges of data security and compliance under the open-source model. Finally, it will look ahead to how building an open-source ecosystem can drive the scaled adoption of trustworthy financial agents and help the financial industry achieve smarter and more secure service upgrades. Outline: 1. Introduction: The Rise of Financial Agents and the Demand for Trust A brief overview of the accelerating penetration of AI in the financial industry. Financial agents, as an important form of application, bring value to scenarios such as risk control, compliance review, and intelligent investment advisory. This section emphasizes the special requirements of “trust” in the financial sector, which must meet core demands such as security, reliability, compliance, and explainability. 2. The Power of Open Source: Supporting the Construction of Trustworthy Financial Agents The advantages of open-source models in technological innovation: open collaboration accelerates model iteration, transparency improves explainability, and community co-creation lowers industry entry barriers. 3. Practical Exploration and Challenges Practical directions for trustworthy financial agents: enhancing information traceability through RAG technology, improving system reliability through modular architecture, and achieving efficient collaboration through the MCP protocol. 4. Challenges in the Open Source Model Issues such as data security and privacy protection (due to the sensitivity of financial data), ensuring model robustness, and risks related to intellectual property and regulatory compliance. 5. Future Outlook: Building an Open Source Ecosystem to Drive Trustworthy Development Looking ahead to a deep integration of open-source technology with financial scenarios, enabling safer and more efficient intelligent services.
