Creative ways hedge funds are using LLMs - Dec 30, 2025
Grading sellside analysts, identifying technical defaults, organizing Iran Notice disclosures and more
Grading sellside analysts
A former sellside analyst used an LLM to analyze 10 years of earnings calls and identify analysts with the most prescient questions. Examples:
Umer Raffat from UBS - Biogen (BIIB) Earnings Call - Oct 2019:
…the implication in the data is that the high with insufficient exposure at the high dose, the second trial worked as well. But when we look at CDR Sum of the Boxes low dose actually looks more consistent than the high dose…how spot-on is that finding on patients that had a sufficient exposure?
LLM reasoning:
Raffat immediately spotted the fragility of the efficacy claim—dependence on a post-hoc exposure subset rather than the pre-specified ITT, foreshadowing the FDA advisory-panel scepticism and eventual withdrawal.
Return since: -20%
Vivek Arya from Bank of America - Broadcom (AVGO) - Dec 2023:
yesterday one of your peers suggested that the market for AI accelerators could be as large as $400 billion... how does Broadcom participate... and what does this larger AI accelerator market imply for your Ethernet networking business?
LLM reasoning:
Arya’s question forced management to quantify AI’s revenue contribution and linkage to networking months before the Street built outsized AI models. The $400B TAM figure sounded extreme in Dec-2023 yet is now consensus; Ethernet’s role in AI back-end networks—dismissed by many at the time—is today a core driver of Broadcom’s upside.
Return since: +338%
My take: Great LLM use case, figuring out which analysts were right for the right reasons. Would be interesting to run something similar on initiations / ratings changes, controlling for sector returns, comparing insights vs peers. Of course, I’d want to know what these analysts are asking about next (here’s the analyst recommendation pages for Umer Raffat and Vivek Arya).
Grading your own analysts
Walleye Capital records all internal conversations to analyze which of their teammembers are most prescient
…we really record every single Zoom, every single call.…So a big part of my job overseeing the risk of the firm, the chief investment officer title, every single morning, me and my risk team, sort of in the control center of running this giant process we have our risk calls and those are all recorded and we can go back and say, hey what were we talking about at this time? And continually have LLMs that are, that are processing those transcripts and helping us to both remember and provide insights and ultimately be a bit predictive, which has been hugely helpful just in that exercise, which is— We haven’t sort of talked about where I think this is going in the power of all this. And I mean, I do believe that we’re that we’re a leader. I don’t want to say we’re the leader because I definitely don’t know what other firms are doing, but I certainly think that we are a bit more advanced in our thinking of how to use these tools, but we’re just scratching the surface of what is possible once you actually start connecting all bits of information within the walls of the firm…
My take: Would recommend an intermediate step focused on returning verbatim excerpts first, folks are going to want a lot of auditability.
Finding short opportunities in bond indentures
LLMs may have identified the opportunity to accelerate Avid Bioscience’s debt and short their stock. Last March, Avid Biosciences received an acceleration notice because it had failed to remove a restrictive legend on its 2026 notes, causing it be in default. Avid shares declined 28% when it revealed it needed to raise $160mm in private placement to redeem the notes.
Byrne Hobart covered this in The Diff, concluding:
And, right now, it’s suddenly gotten much easier to do this at scale: you can unleash LLMs on indenture agreements, and try to find edge cases that the company didn’t think of or notice. These will all be technicalities in practice; in the Avid case, if the restriction had been a big deal to the note owner, they probably would have noticed right away. But, perhaps coincidentally, they only noticed after the newly-widespread availability of tools that can trawl through vast amounts of text to extract useful information.
In Money Stuff, Matt Levine weighed in that he was skeptical an LLM found this opportunity.
My take: Hard to be sure, but I think this opportunity *was* found by an LLM because the acceleration notice was received just two weeks after Google released Gemini 1.5 Pro - the first time a 1mm token context window was generally available - enabling the analysis of huge documents. Would have been straightforward for any fund to cycle through indentures to identify technicalities that could merit an acceleration notice. In fact, this would make a pretty interesting eval for new models that get released: run them against a huge corpus of indenture agreements and see what new opportunities get identified.
Iran Notice disclosures
John Friedman, CEO of Datamule, collected and published a searchable dataset of Iran Notice disclosures from SEC filings
My take: one way I think about LLMs is they enable instant creation of datasets that are plausibly interesting but not worth hiring and waiting for a human team to build.
Interesting job posts
Select buyside LLM related job descriptions:
Point72 - AI Engineer – Investment Research & Workflows ($150K-$200K)
This role partners directly with L/S equity portfolio managers, analysts, and business leadership to build innovative solutions to improve efficiency and research quality across the equities platform…
Longaeva (new Baly platform) - Research Product Associate - AI Enablement
Longaeva is adding an associate to join the proprietary research team to accelerate adoption of generative AI products across investment strategies. In this role, you will embed directly with the proprietary research and investment teams to build solutions that impact investment decisions. We are seeking a capable, technical candidate—someone able to do hands-on research product development, web scraping, and LLM/AI-powered synthesis of qualitative and quantitative data. The ideal candidate blends scrappy coding, data/information aggregation, and a strong product intuition, with a proven ability to ship projects fast and independently. You will translate our AI capabilities into actionable insights by rapidly prototyping agentic workflows, building novel research products, and driving adoption of in-house tools.
Bayview ($30b AUM credit firm) - LLM Analyst ($90K - $110K)
The Research team at Bayview Asset Management is hiring an LLM Analyst to unlock insight from large volumes of textual data, both external and internal, to inform investment theses, improve operations, and answer foundational questions about the mortgage industry and more broadly, the economy...Meet with portfolio managers, traders, marketing and servicing teams to identify and narrow down the question. Understand the business context behind each question….
Prototype quickly but evaluate rigorously: Design prompts, run experiments in notebooks and concisely synthesize results for fast iteration. Define clear success metrics to measure progress.
xAI - AI Buy-Side Finance Tutor ($45/hr)
We are seeking a skilled AI Buy-Side Finance Data Specialist to enhance xAI’s AI models by providing high-quality data annotations and inputs tailored to buy-side finance contexts. In this role, you will leverage your expertise in portfolio management, hedge fund strategies, private equity investments, venture capital deal sourcing, and high-frequency trading algorithms to support the training of AI systems. You will collaborate with technical teams to refine annotation tools and curate impactful data, ensuring our models effectively capture real-world buy-side finance dynamics.
My Take: Interesting these are mainly early career type hires, no graduate degree required. Looks like funds are happy to build on top of frontier models and popular tools. Lots of focus on prototyping and experimentation. The Grok LLM trainer hourly rate feels a little light!!
Follow for more case studies of LLMs+Investment Research
If you would like to discuss incorporating LLMs into your research process, reply to this email or reach out via X or LinkedIn.