Hong Kong Machine Learning Season 7 Episode 2
24.04.2025 - Hong Kong Machine Learning - ~3 Minutes
When?
- Thursday, April 24, 2025 from 6:00 PM to 8:00 PM (Hong Kong Time)
Where?
- This HKML Meetup is hosted at Maven Securities, Wan Chai, in Hong Kong.
The page of the event on Meetup: HKML S7E2
This event was organised by Vahid Asghari.
Programme:
Talk 1: MAVRAG: Multi-Agent Vectorless RAG
Abstract: Learn from trial and error about answering questions in natural language by querying multiple data sources without stale data using a multi-agent hierarchy within the RAG framework.
Short bio: Alex Hunsberger is a platform engineer at Maven Securities, and has also been a developer for 10+ years with interests ranging from Python to Linux to DevOps and recently Machine Learning.
Talk 2: Investing in a complex world
Abstract: In this talk, Emmer Capital CEO Manishi Raychaudhuri outlines a contrarian macro and equity outlook for 2025, arguing that Asia is poised to outperform the United States amid shifting global economic dynamics. The presentation explores the erosion of U.S. exceptionalism, the nuances of recession risk, and the implications of a depreciating dollar. Emphasis is placed on Asia’s relative macro resilience—supported by fiscal and monetary policy space, undervalued equity markets, and improving earnings revisions. The talk culminates in a “Quantamental” stock screen highlighting high-growth, high-ROE Asian equities trading at attractive valuations, with a focus on domestically exposed sectors and quality growth at reasonable price (GARP) names, especially in China and Hong Kong.
Short bio: Former APAC Head of Equity Research at BNP Paribas, with 27 years in sell-side research. Manishi Raychaudhuri led teams, built market-beating strategies, and boosted research quality. Ranked top in Asiamoney/II polls at UBS India. Expert in Asia-Pacific equities, investment themes, and corporate access. Frequent media commentator and speaker. Began career at ICICI Securities (JP Morgan JV).
Talk 3: Risk Monitoring for Financial Firms: An Application with Machine Learning Methods
Abstract: In this talk, I will explore how financial firms’ risks can be monitored using accessible machine learning methods. Emphasizing practicality over cutting-edge complexity, I will demonstrate how we developed a robust empirical framework for studying the interconnectedness of financial institutions and its evolution over time. Using only daily volatility data, which is readily available to practitioners, we estimate total connectivity through a vector autoregressive (VAR) model, enhanced with factor-adjusted regularization techniques. I will highlight key insights, including how global financial connectivity has responded to systemic risk events. Finally, we will discuss how this approach might extend beyond systemic risk measurement.
Short bio: Johnson (Juncheng) Li is a Ph.D. in Business Statistics at The Hong Kong University of Science and Technology, specializing in financial econometrics and high-dimensional data analysis. Johnson has experience processing over high-frequency financial data and developing statistical models for risk assessment. Before his postgraduate studies, he led a team to 4th place among 150 teams in the Asian Supercomputer Contest. With his combination of strong academic credentials and practical industry experience, Johnson brings valuable expertise to data analysis and financial risk modeling.