A day in the life of an applied AI engineer at hedge fund Balyasny
Based in Austin Texas, Michal Mucha is a lead engineer on Balyasny’s Applied AI team. He’s worked for the hedge fund since 2022 and has been part of the Applied AI team since its inauguration last November. This is what a typical day looks like for him.
Get Morning Coffee ☕ in your inbox. Sign up here.
6.30am. I have a young family. My kids often wake me up early and we’ll spend some time having breakfast together before I leave for work.
7.45am. I usually drive to work. It takes around 15 minutes. Sometimes I’ll come in a little later if I’ve had a late-night coding or I want to drop my kids off first. The office is very well positioned and easy to get to. We have a strong AI presence here in Texas – I’m working with Peter Anderson, our head of research in applied AI who joined from Google earlier this year.
8.30am. We usually have quite a few meetings in the morning. I’ll meet both within our team and with the broader investment and business teams using our tools.
Our remit is fundamentally to create tools that can – ideally – be scaled for use by our investment teams across a number of different asset classes and that will help to improve the lives of our users. Our tools are already used by 80% of our organization and we’re very proud of that. Almost all the work I do involves large language models (LLMS).
When we meet as the AI team, it will often be to discuss research papers or experiments we’re designing and running. We try to keep up with research papers released by key players and often run a lot of different experiments concurrently.
9am. I have a meeting with some analysts who are interested in using one of our tools. It’s a kind of bootcamp and an opportunity to get their feedback on their experience with it.
10am. I'm preparing for a firmwide demo of our latest release of our proprietary ChatGPT-like system, which leverages internal resources and proprietary datasets. I led the development of the initial version, getting it up and running in just 4 weeks. It's an exciting but complex product, as we've had to navigate stringent compliance and security requirements while retaining its core capabilities.
11am. I join head of Applied AI for a meeting with some portfolio managers. BAM GPT is divided into different “assistants” for different asset classes – we have a flagship set of assistants for areas like equities, and we’re always rolling out improvements to them and looking at expanding the range available to investment teams. Meetings with portfolio managers are an opportunity to establish where our gaps are. We try to go for a teach the teacher approach and we have power users in each team who work very closely with us.
12pm. I usually have lunch around midday. Mostly we eat in the office, but sometimes we’ll go to a nearby restaurant. I’ll often take the opportunity to go the gym at lunch too – I find that lifting some weights in the middle of the day gives me clarity in the afternoons.
1pm. After lunch, I usually spend a lot of time with our experiments. A lot of what we do is very innovative, and BAM GPT has become so good that it’s become my preferred LLM even when I’m at home.
The experiments are typically structured around achieving a business objective. We’ll usually start by framing what the benefit to the user will be and we’ll then evaluate how we can achieve that.
Some of our recent experiments have been undertaken with Chris Pulman, our head of Macro research in London. He used to spend two days preparing for Fed and central bank previews, but he brought his process to us, and together, we were able to get that down to 30 minutes by plugging part of his analysis into BAM GPT.
3pm. We have a release. We produce a velocity of work and have been able to move very fast. We also have a great testing environment. Often someone in the AI team will have a great idea from a conversation with a portfolio manager and we’re able to turn it around very quickly. We might easily have several releases in a week. They’re often about rolling out improvements to one of our existing assistants.
4pm. We have another meeting with some users. These meetings are often about asking the right questions and understanding how tasks can be broken down into smaller pieces. Getting this right is critical to automating workflows.
5pm. Once a month, we have a research club where we take turns presenting novel research papers to the AI group. We’re always trying to anticipate what’s coming next, and the research papers will often determine the direction of our experiments.
6pm. I arrive back home and there’s a pause in my life for the family. The kids have got into riding bikes and we’ll go out or watch movies and play Minecraft. When everyone’s asleep, though, I’ll often start coding again. I’m a bit of a night owl and I love coding at night – I like the peace of mind that comes from knowing no one will interrupt me, and some of my best work is done later in the day.
Have a confidential story, tip, or comment you’d like to share? Contact: +44 7537 182250 (SMS, WhatsApp or voicemail). Telegram: @SarahButcher. Click here to fill in our anonymous form, or email editortips@efinancialcareers.com. Signal also available.
Bear with us if you leave a comment at the bottom of this article: all our comments are moderated by human beings. Sometimes these humans might be asleep, or away from their desks, so it may take a while for your comment to appear. Eventually it will – unless it’s offensive or libelous (in which case it won’t.)
