The asset management industry is evolving rapidly, with the introduction of machine learning models, the rise of alternative datasets, and advanced data extraction and analysis. These advances are also raising new challenges (interpretability, overfitting, sample selection biases, systemic complexity, small sample issues, etc.).

The seminar “Quantitative Management in the Machine Age” organized by Abu Dhabi Investment Authority and Institut Louis Bachelier, offered the unique opportunity to discuss some of these challenges with key experts in the field.

Recent academic developments are offering attractive solutions to some of these issues. For example, imposing some economic structure helps make forecasts more accurate, robust, and interpretable. It can also provide novel ways to obtain factor risk premiums directly from derivatives prices. Portfolio allocations can directly modeled as a function of signals, which removes some of the complexity inherent in traditional portfolio management approaches.

Data analysis has become so important nowadays that several teams and entire research areas have been built just around the ideas of dataset setup, data curation and features extraction. Asset managers are using those data to run prediction models, or extract causal relationships within the data.

In practice, when making fast real time data-based decisions, one need to incorporate the fact that the environment is subject to structural changes. Continuous monitoring of the decision making process with rigorous tests, having a human in the loop, and training and testing algorithms under various scenarios of non-stationarity can help. A second challenge concerns the creation and management of algorithms by data and analytics teams, scaling, and acknowledging that performance might drop suddenly. The tech sector has great experience in doing that for large scale platforms, MLops being particularly useful, and therefore should be observed closely.

At the organization level, investment firms often hire specialists (e.g., an expert in NLP), but ask them to work in silos, where they rise through the ranks by becoming generalists (e.g., portfolio managers). Under this silo and platform structure, quants face the challenge of beating the collective wisdom of the crowds. A research lab structure offers a unique environment to develop sophisticated strategies that are beyond the grasp of silo/platform firms, by means of: (a) co-specialization, working in a highly collaborative environment; (b) tackling well-defined open investment problems; and (c) applying scientific methods. This is a promising avenue for quantitative asset management in the industry.



Open Problems in Quantitative Investment
Marcos Lopez de Prado, Global Head of Quantitative R&D, Abu Dhabi Investment Authority, ADIA and Professor of Practice at Cornell University

Creating Alpha with Machine Learning, some examples
Gautier Marti, Quantitative R&D Developer, Abu Dhabi Investment Authority, ADIA

Data Curation: a prerequisite for Systematic Investment?
Charles-Albert Lehalle, Global Head of Quantitative R&D, Abu Dhabi Investment Authority, ADIA and Visiting Professor at Imperial College London