Sponsored Content
AI Will Help CFM Discover New Sources of Alpha
The convergence of AI, data, and human expertise is reshaping alpha generation.
September 15, 2026
Sponsored by
CFM
For Capital Fund Management (CFM) data science and the latest machine learning and AI <a href="https://myappsplus.com/as-the-world-debates-the-risks-of-ai-china-closes-the-technology-gap-with-the-us/” title=”As the world debates the risks of AI, China closes the technology gap with the US”>technology are core to finding new sources of alpha, speeding up time to investment processes and enabling efficient scaling across other asset classes.
The quantitative and systematic investment firm utilizes the latest techniques to develop alternative investment strategies. In doing so, it strives to remain on top of technological advancements as they evolve.
Institutional Investor spoke to CFM’s chief product officer Laurent Laloux and chief technology officer Benjamin Roy about how the firm is using cutting edge data-driven technology to improve the service it offers investors.
Institutional Investor: How has the integration of machine learning, AI and other advanced data science techniques improved the investment process and helped CFM discover new sources of alpha?
When I started 30 years ago, stock exchanges and futures exchanges provided good quality data that allowed us to use modern statistical techniques to develop strategies. More exchanges, new data providers with alternative data sets, easier access to computing power , and modern user friendly technology stacks have since helped quant researchers surf the wave of technology to provide newer sources of alpha and to refine it more efficiently.
Is the S&P 500’s Concentrated Rally Starting to Diversify?
Sponsored by CME GroupSeptember 7, 2026
AI industrializes our discovery of data sources. Previously, alpha selection was driven by human beings making sense of the data, building an investment thesis idea and validating the back-test results. Going forward, thanks to AI, we’re industrializing this process at scale and we’ll generate anywhere between 10 to 100 times more models. The next challenge is to develop an industrial methodology to assess and select the most robust models at this scale.
Benjamin Roy
One thing that those techniques have helped us change is our research velocity. We have built the CFM platform leveraging machine learning, AI and data science techniques. It enables our researchers to be more creative, productive, and rigorous in how they think about alpha.
Having that platform helps us test more ideas and get better signals, and in the end better performance. It has really compressed the time it takes on the path from an idea or hypothesis back test to production.
Given the abundance of alternative data today, how does CFM determine what data has genuine power of predictability versus just simple noise?
Laurent Laloux
We’ve been working on this question for more than 10 years now. With the emergence of all these alternative data sets it is important to have a full vetting process for new data. That starts with data sourcing and assessing the quality of the provider.
When a provider has good legal documents and Due Diligence Questionnaires, good quality data tends to follow that we can use to build models. It is important to spend some time digging into the data, removing spurious effects.
We then have various specialist teams assess the data set to ensure it makes sense across different angles, before doing a basic test of alpha.
If it all looks interesting after that then we’re ready to really dig into the data. You can detect trivial effects easily with modern machine learning techniques. These are the ones that tend to decay faster, because all our competitors – whether prop traders, HFTs, or quant hedge funds – will also find them easily and jump on it. Our edge is finding more complex and high quality data sets.
We spend a lot of time digging into data without asking for the data set to be profitable in one month or even three months. We are willing to explore a data set for a very long time so that we can extract the best value that we can for our investors.
What role do more advanced AI techniques play today in the investment and research process compared with the classical statistical methods?
Laurent Laloux
To control quant research you must tame in-sample bias problems as much as possible. In the past that meant asking a human being with a prior statistical, technical, or economic model or idea to confirm and then build a model that makes sense.
This can now be done by an agent, provided you script their skills and constraints. As a quant researcher, you really have to take a step back and ask what your job and thought process is all about, and how you can code that into an AI agent model. If you do that correctly, instead of spending time generating one model you can have an agent producing hundreds of models.
Now the question and the problem you have is how to select and vet these 100 models.
The agent will be extremely good at finding all the minute mistakes in the data, at leveraging all the spurious correlations, and at generating models that look beautiful on paper but are totally bogus out of sample. The question is how to ensure you select the best models while managing in sample bias on an industrial scale.
Looking ahead, will CFM be able to trust AI to such an extreme that you can reduce headcount and oversight capacity so that in say five, 10 or 25 years, you’d be half the size?
At CFM we see AI as a productivity and efficiency tool, not a means to reduce head count. Part of the technology team’s work at CFM is to think about how we build processes to increase the focus and reliability of the models that are produced. We are also leveraging technology to enhance, improve, and scale up reliability controls on models. Human oversight is still key in our process, both on the research side and on the tech side.
AI has allowed us to compress the research process. By giving it less value-added tasks it frees up even more time for a researcher to focus on the reliability of the data or the time spent on a specific data set to dig into specific features or to build new ones.
Over the past few years at CFM, the more data we have and the better our data is, the better the process is. It is really the core of our business and what we’ve been doing for decades. The more technology progresses, the more that big flywheel compounds.
Laurent Laloux
I don’t think we’ll reduce the size, frankly. We’ll make sure that the people we hire have the right skills. But the core is that it’s only an engine. If your data is not properly curated and of high quality, it doesn’t serve any purpose. The focus on data quality, data curation and features generation is even higher than before because of that.
You cannot escape that, and it still requires a lot of human expertise to make sure that your data is properly created, properly available, and easily accessible.
You can industrialize the day-to-day job of quant research, but we still need a lot of bright people to look for the next evolution or edgy technological, mathematical or statistical breakthrough, which might change the future.
CFM trades across multiple asset classes, how do data science platforms enable signals or models developed in one market to be efficiently adapted to others?
Laurent Laloux
I guess the key is for us is that people are not siloed. We have an internal data catalogue which is open to all data scientists and researchers. Everybody can look at what’s available with explanation and content.
When we test data we like to have a different specialized team look at it from macro and micro angles, or from execution or high frequency signal angles, so that they can share ideas. The technology team can assess what a model that has been built for futures can do on equities, and vice versa.
Each time a model is generated they make sure it can be vetted across different asset classes easily. That’s essential. If a model has been designed with one asset class in mind and it works across others, there is an even stronger case that it will provide robust alpha for the future.
What edge does CFM have over the rest of the market?
Laurent Laloux
We are proud of our culture, which we’ve managed to build and keep alive since day one. We are collaborative, always ask questions and never assume that what we have is the best. There’s always something new and better to do. We always make sure we are there listening to the latest technology, statistics or methods that might change our business.
Having a deeply research-driven collaborative culture is something that we feel really makes us who we are. I’m not claiming that we are the only hedge fund like that, there are probably a handful, but it is something that we feel makes us strong, gives us an edge in finding and maintaining alpha and helps us deliver better outcomes for our investors through our funds.
Any description or information involving modes, investment purposes or allocations is provided for illustrative purposes only and does not constitute investment advice nor an offer or solicitation to subscribe for any security or interest. Any statements regarding correlations or modes or other similar behaviors constitute only subjective views, are based upon reasonable expectations or beliefs, and should not be relied on. All statements herein are subject to change due to a variety of factors including fluctuating market conditions and involve inherent risks and uncertainties both generic and specific, many of which cannot be predicted or quantified and are beyond CFM’s control. Future evidence and actual results or performance could differ materially from the information set forth in, contemplated by or underlying the statements herein. CFM accepts no liability for any inaccurate, incomplete or omitted information of any kind or any losses caused by using this information. CFM does not give any representation or warranty as to the reliability or accuracy of the information contained in this document.
