SESAMm collaborates with Kyobo AXA IM to construct & implement quant strategies

by: Jorge Alvarez , 2 minute read , October 4 2022

We’re happy to announce our collaboration with Kyobo AXA IM in the construction of machine learning models based on SESAMm’s NLP alternative data extracted from the web to process and read more data.

“Since 2020, Kyobo AXA IM leveraged SESAMm's AI technologies to build smarter quantitative strategies, including asset allocation and trading,” said Frédéric Bellemin, Deputy CEO,” As a result of SESAMm's alternative data and NLP capabilities, Kyobo AXA IM has generated superior returns, saved time, and further integrated machine learning techniques into their investment processes.”

Over the last few months, we’ve worked with Kyobo AXA IM’s team to implement two financial use cases and quant strategies leveraging SESAMm’s TextReveal® natural language processing models.

“We’re very excited about our partnership with Kyobo AXA IM and particularly about the results of our work together,” said Sylvain Forté, SESAMm’s CEO & Co-founder. “It’s gratifying to see how SESAMm’s technology adds significant value to quantitative strategies everywhere in the world, in this case working on South Korean equities and other asset classes.”

Asset allocation (optimal AA weights)

We started by creating a global trading universe divided into 3 asset classes: equity (the US, world ex-US, EM), bonds (US treasury, corporate), and alternative (commodity and REIT). Then, we and Kyobo AXA’s team created two signals to detect asset classes with the most upside and downside potential. The first signal is benchmarking, made with widely available financial and market data. The second signal leverages NLP features to capture intangible information unavailable in financial data. We also extracted the best trading signal possible by combining all features with a machine learning algorithm. 

Each signal is traded with traditional risk limits and realistic trading constraints. We substantially increased the Sharpe ratio of the final strategy using NLP compared to the strategy without NLP data.

Equity trading

For this use case, we went through the same process as the previous one. We created two signals to distinguish over and under-performing stocks of the KOSPI 50 index. The first is benchmarking, and the second is NLP-fueled. 

We executed this use case using NLP and market data to find over and under-performing stocks of the KOSPI 50 Index, which allowed us to create a trading strategy that significantly increased the performance on a risk-adjusted basis compared to previous strategies and benchmarks. Both signals are traded with a traditional quantitative portfolio model with realistic trading constraints. This second strategy outperformed the first one by a double-digit percentage on a risk-adjusted basis.

This collaboration not only delivered higher returns, but the team also learned so much from it and got a grasp of SESAMm’s technology. These projects will open up the path to more long-term collaborations to better serve Kyobo AXA IM’s clients by leveraging SESAMm’s cutting-edge technology.

Contact us

To learn more about the two use cases, contact us at or