blue texture background

AI: Not completely new to financial markets but continuously exciting, says Wachi Bandara, DRW’s Head of AI

Wachi Bandara is DRW’s Head of Artificial Intelligence (AI). He has spent his career solving problems at the intersection of trading, machine learning and data science. A successful founder with several exits, he’s built and led multi-disciplinary teams working towards solving problems with significant real-world implications. In this discussion, he talks about how markets have used AI and ML for decades, how it makes markets more efficient and why DRW is an attracting top AI talent.

How did you get into finance?

My background is in math and computer science. I started my career doing extensive research on image processing and computer vision algorithms. During that process, I discovered trading and finance. After grad school, I went to work for Mellon Capital focused on yield curve arbitrage and equity risk for the firm’s global macro fund.

After moving to San Francisco, I saw an opportunity to apply Machine Learning (a specific type of AI) to unstructured data and difficult data sets to transform them into insights that are useful for a trading and investment use case. So, that became my focus and I co-founded a couple of startups in the space. Don and I met during my last exit, and I joined DRW in 2023, based in Silicon Valley.

You were an “early adopter” of AI – how has that influenced how you think about the technology?

If you never followed or paid attention to this space, then it looked like a sea of change in 2020 when Open AI introduced GPT-3. I do think the transformer architecture is revolutionary and OpenAI and others have made AI more accessible to the general public in exciting ways, however, computational linguistics has been around for a long time and we have used sophisticated algorithms that process language successfully in markets for decades. I see the developments of the last couple of years as the next step in a long chain of advancements that have evolved the technology.

I think being in the space for a long time gives me a more nuanced perspective that fits well with DRW’s approach. I really enjoy the work of asking a lot of questions, doing a lot of testing and understanding what drives these models and how they relate to the problems we are trying to solve. My goal is to be very thoughtful and understand why things work the way they do, which I think ultimately yields the best outcomes.

How are you using AI and how is it evolving?

Generally, when people think about AI, they think about LLMs (Large Language Models). And while LLMs are certainly exciting, the things we’ve learned from LLMs have many other applications. My team’s work is focused on understanding what those use cases could be and then to make sure the firm is not only using the technology appropriately, but helping to shape it in a way that is responsible and sustainable.

What’s most exciting to me at DRW is that it’s an opportunity to take this technology and marry it to things that are specific to DRW’s competitive edge. I think that’s where we’ll see the real power of this technology as it continues to evolve -- right now, we're using it to solve general problems but I think many industries, from medical to climate to markets – will see tremendous value as we couple this technology with very specific information to create very precise and targeted outcomes. I think that’s what I’m most excited about – working at a firm where my mandate is to build the most innovative products in the world in a way that is thoughtful and makes it a sustainable model for the next decade.

How will AI change access to financial markets?

In some ways, I think that change has already happened. When you look at firms like DRW and other institutional participants in the market, they’ve been deploying models for a long time, and what we’ve seen is that has driven down trading costs and democratized markets over the last few decades. For us, this next generation of AI is about being thoughtful in continuing that work and building new tools and solutions that keep us on the cutting-edge.

I do think we’re just starting to see the use cases for AI on a more individual investor level, which is interesting. Just ten years ago, you had to be a high net worth individual to get access to certain types of investments – or even information. Now, incremental costs of doing business are much lower and as a result we’re already seeing more startups that offer investment opportunities to previously marginalized groups. I think we have to ensure that this technology is harnessed appropriately in those uses cases to benefit those investors.

What’s being done to manage AI risk? How is it evolving?

In theory, AI has paradigm-shifting possibilities. In reality, those possibilities have to be squared against the resources and discipline required to use it appropriately.

AI models are powerful technology with known weaknesses; they’re good at qualitative answers and poor at quantitative answers – at least for now. You experience this first-hand this when you experiment with ChatGPT. We’re aware of the hallucination problem in AI and we try to minimize the impact. It reminds me of some of the pushback we saw with algorithmic trading in its early days – people believed that it was just machines running with no human oversight. There were always humans ultimately holding the controls and validating the work there, and we’ll need the same with AI as its deployed; or at least understand what is analogous.

Collectively, we also must pay attention to how we feed and train models to reduce risk and avoid bias. As more data becomes available, we can improve models. The world is generating more data than ever before: A report from a few years ago estimated that 90% of the world’s data has been generated in the last two years alone.1 The proliferation of more data means that testing is easier, and markets should leverage data to ensure even more rigorous testing.

How far are we on the AI disruption scale?

AI is just getting started. The world will be dramatically different in 20 years. Today, we cannot live without our smartphones (Siri, by the way, is a great example of AI already in use) and the internet. The next generation of AI tools will likely become just as familiar to us in our daily use.

In fact, the technology is so different and powerful that we need a new paradigm to think about what success is. AI is already creating a set of startups across hardware, software and applications that will reshape society in the next decade.

Why should AI talent consider DRW?

We’re a small team doing very innovative work in AI. We’re a good fit for people who want to be at the forefront of AI research, doing cutting edge work that is not just academic. At DRW, we’re committed to collaboration and challenging consensus, and so the people who thrive here tend to want to have the freedom to think critically and ask questions about what’s possible. We also have high standards – our goal is to deliver excellence in our outcomes, not just for the firm but for the markets overall.

We’re committed to growing our AI team and positioning ourselves as a destination for innovative minds. Nimble firms have advantage in AI, especially when it comes to scaling and implementation. DRW is an exciting place to be for AI specialists because we get to test and iterate quickly. As Don says, there’s great value in a fast feedback loop.

(1) Retrieved from: