When AI meets quantitative finance: A PhD candidate seizes the opportunity

AIFMRM PhD candidate Freddy Dipudi has been using artificial intelligence tools for a long time. He knows their power – but also their shortcomings. That’s why he’s building a generative pre-trained transformer, or GPT, tailored to quantitative finance. We recently sat down with Freddy to hear about his latest project.

Artificial intelligence is changing how we work and live. And it’s changing how we learn.

While GPTs – generative pre-trained transformers – can give students access to a wide range of information, they’re generalist by design. That means that more often than not, they don’t have instantly accessible, cutting-edge data in a particular field. But this is the kind of data you need if, like Freddy Dipudi, you’re pursuing a PhD in quantitative finance.

“I realised in my own work that GPTs are lacking when it comes to offering the latest quantitative finance research,” Freddy says. “But they’re extremely powerful tools. And here’s the thing: you can fine-tune them for specific purposes.”

So, for the last month, Freddy and AIFMRM Adjunct Associate Professor Tanja Tippett have been tailoring a GPT to meet the needs of quantitative finance students. Alongside his close work with Tanja, Freddy’s had continuous support from AIFMRM Adjunct Professor Tom McWalter, Associate Professor Peter Ouwehand and Senior Lecturer Dr Khouzeima Moutanabbir.

Currently in its pilot phase, the GPT is being built specifically for AIFMRM’s Quantitative Methods of Finance course. Should the pilot be successful, the team will expand the GPT to other major courses. The team is using ChatGPT model version 5 as a base, because it is already ahead of other large language models (LLMs) when it comes to mathematics and coding. Freddy, Tanja, Khouzeima, Tom and Peter are taking it even further – and they have good reason to.

“Imagine you’re at university,” Freddy says. “It’s two am, you’re studying, and you have a question about the problem you’re currently working on. Regular search engines and LLMs won’t have the answer. Your tutor might know what to do, but there’s no way you could, or should, contact them at this hour.”

That’s where the quant finance GPT comes in. Because it’s trained by experts, it’s able to provide expert-level guidance at any point in time, at any scale.

But it’s not there to replace tutors and researchers. “It’s a tool,” Freddy says. And it’s only that: something you use in order to make something else. 

Freddy is well aware that students are using similar tools already. “But we just want to make sure we give them a safe, accurate, and more effective GPT that they can use to learn and that they can trust,” he adds. 

“AIFMRM always wants to give every student the best experience. The question the GPT team is always asking is: How can we leverage this new technology to improve this experience even more?” 

Freddy, Tanja, Khouzeima, Tom, Peter and everyone else at AIFMRM are sure to find the answer.