Real People, Real Impact: How Alternative Data Powers Investment Decisions at Citadel

What does DSG do and how does it support how Citadel generates alpha?

DSG turns complex alternative datasets into clear, decision-ready insights for our Investment Professionals (IPs).

We focus exclusively on alternative data rather than traditional market data. Alternative datasets are large and messy. Real-world and real-time datasets, like satellite images, can be very valuable for investment insights, but challenging to work with directly.

Our job is to translate raw, often noisy data into outputs that are intuitive, timely and directly usable. These insights inform real investment decisions. For example, it is well-known that some hedge funds analyze satellite images of retailers’ parking lots to predict demand. Our research follows similar examples but uses broader varieties of real-time and real-world data. We use statistical, AI-based and machine learning methods to process and model such datasets.

Better information, delivered earlier and with clarity, helps Citadel generate alpha. We are a central quantitative research team working with all investment teams across the firm— including fundamental and quantitative investors across equities, commodities, fixed income and macro—so that we can help Citadel maximize commercial value out of alternative data sources.

As a quantitative researcher, what do you do at DSG?

As a quantitative researcher (QR) at DSG, my role has three components:

1. Partnering with investment professionals to understand the decisions they are trying to make and the hypotheses they want to test.
2. Identifying and evaluating new alternative data sources to check which ones are additive to what Citadel already has in-house. Given our experience, we have quick turnaround times, and we are comprehensive because we know which use cases are deployed across the firm. We evaluate data not just to answer immediate IP questions, but also to uncover new types of analyses that IPs may not yet have thought about.
3. Extracting signal from the data by applying statistical, AI-based and machine-learning methods to convert raw inputs into actionable insights.

Every project is different. Each dataset comes with its own structure, noise and constraints. That means, the work is not about reusing the same playbook; it’s about learning quickly, adapting and innovating methods, building new capabilities, solving new problems and staying at the bleeding edge of testing the latest AI / machine learning applications to create commercial impact.

What types of data do you work with?

I work with a wide range of alternative datasets. Besides satellite images, for example, I also work with structured data like prices and unstructured data like text. Citadel is deeply committed to giving researchers access to cutting-edge data sources so we can solve challenging problems better and faster than our competitors.

The work is grounded in real behavior. I’m constantly asking, What is actually happening in the world, how does it appear in data and why might it matter for markets?

I often describe the work as a treasure hunt. IPs bring questions and hypotheses, and I think creatively about:

• What data could help answer those questions
• How that data acts as a proxy for real-world activity
• What methods best extract a reliable signal

That blend of intuition, creativity and statistical rigor defines how we approach alternative data.

What does a typical day at work look like?

Most days, more than 70% of my time is spent on active research – analyzing data, building models and iterating on signals.

The rest of the day is collaborative, meeting with IPs to align priorities and discuss our latest capabilities and alternative data sources, and meeting with teammates to challenge each other’s research assumptions.

Projects follow a clear lifecycle: sourcing new alternative data, conducting research, working directly with investors to shape commercial use cases and partnering with quantitative developers to bring successful ideas into production. Unlike in academia where each iteration could take months, we often see commercial impact within days or weeks of deploying new research.

What is your background? And the backgrounds of people you work with every day?

I spent more than eight years in academia working in astrophysics and cosmology. My research involved analyzing massive, noisy datasets to study dark energy and dark matter and developing statistical techniques to maximize the information extracted from them.

While the subject matter has changed, the core skills have not: statistical reasoning, large-scale data analysis and research-driven problem solving translate directly to DSG.

Across the team, DSG brings together people from a wide range of STEM disciplines. Some come from theoretical fields like mathematics or physics; others from engineering, computer science, or applied sciences. This diversity is a real advantage. Different training produces different ways of thinking, and that consistently produces stronger results than any single approach could on its own. We have a very collaborative culture of frequently exchanging ideas across the team, learning from each other and organically forming groups to solve complex problems together.

A finance background is not required. What matters is strong research instinct, curiosity and the ability to learn quickly.

What’s the difference between being a quantitative researcher and a quantitative developer at DSG?

Quantitative researchers focus on research and signal generation: developing ideas, testing hypotheses and validating insights.

Quantitative developers focus on production and scalability: building efficient workflows that can scale across investment teams and investment strategies, deploying research to production with greater reliability and speed and developing new machine learning tools to accelerate research.

The roles work closely together and evolve as a project progresses. Early on, research is QR-led. As the work matures, QDs take increasing ownership to ensure the output is robust, scalable and ready for ongoing monetization. Leadership shifts with the lifecycle of the research problem.

How do you work with other quantitative researchers and quantitative developers in DSG?

Each QR typically owns one or more data modalities, which means leading research, continuously finding ways to upgrade data quality and acting as the subject-matter expert in that domain. This creates clear ownership and real impact for every team member.

While ownership is clear, collaboration is constant. We routinely discuss open research questions internally within DSG and with collaborators across the firm, review results and test new approaches and challenge each other’s assumptions. This happens through structured research discussions held twice a week, as well as frequent, informal whiteboarding and day-to-day conversations.

Every month, we also host “brown bag” talks during which team members present any topics they are passionate about. Topics from the past include quantum computing, chasing cosmological neutrinos in the universe and hiking in the Himalayas. It is an interesting and fun way to learn about your co-workers’ passions and research backgrounds.

How do you work with other teams across Citadel?

We are a central quantitative research team, meaning that we work with all investment teams and investment strategies across the firm. Our most frequent collaborators include fundamental and quantitative investors across equities, commodities, fixed income and macro.

We partner with these investment teams to identify and evaluate new alternative data sources, design research frameworks and deploy new quantitative research that enhances their investment process and generates alpha.

This collaboration gives us a unique perspective on how different investors think about risk, alpha generation and portfolio construction. It’s a great way to learn about diverse asset classes, trading horizons and investor priorities across sectors.