Dieser Artikel ist auch auf Deutsch verfügbar.
Quantitative finance is not one job, and it is not a collection of secret formulas. It is a way of working: turn a question about markets into something that can be investigated with data, mathematics, and code. The result may be a forecast, but it does not have to be. Quants also calculate prices, measure risk, and build systems that execute decisions reliably.
In one sentence:
Quantitative finance combines financial markets with statistics, mathematics, computer science, and scientific testing to make decisions under uncertainty.
One field, four kinds of problem
Titles differ between firms and the boundaries overlap. Four roles still show how different quantitative work can look in practice.
Quant researcher: Turn an observation into a testable idea
Suppose you want to detect changes in global oil supply before a monthly report appears. A researcher might discover a new measurement in satellite images of floating-roof tanks. The roof rests on the oil and moves with the fill level; combined with the sun angle and tank geometry, its shadow can reveal its height. A remote-sensing study shows how that measurement can work in practice.
The creative leap is only the beginning. Which tank types are visible? When was the image captured? Does cloud cover bias the sample? Is the estimate actually available earlier, and does it explain anything in the market? Citadel's alternative-data team describes this translation directly: turn noisy data into information that is intuitive, timely, and useful for a decision.
Quant trader: Decide before certainty exists
A market maker quotes buy and sell prices at the same time. Then a large order arrives, related markets move, and the firm's inventory grows. The trader has to decide: move the prices, reduce the risk, or deliberately hold it? The information is incomplete and there is rarely time for a long calculation.
The work rewards probabilistic thinking, fast estimation, market intuition, and the ability to recover immediately from a mistake. Optiver describes trading in similar terms: decision-making under uncertainty in a fast, adversarial environment, tightly connected to research and engineering.
Quant developer: Make a good idea work in reality
A model can be correct in a notebook and useless in a live system. Market data arrive out of order, a connection stalls, or a calculation becomes too slow exactly when the market gets busy. Quant developers build data pipelines, simulations, and trading systems that stay fast, observable, and controlled when something fails.
This calls for algorithms, operating systems, networking, profiling, and clean interfaces. Creativity often means making a complicated data path radically simpler. At Hudson River Trading, for example, statistical models run directly on high-performance trading infrastructure and are continuously analysed there.
Portfolio or risk quant: Find the failure before the market does
Five positions can look different on paper and still depend on the same hidden risk. A risk quant therefore looks beyond the most likely path and constructs uncomfortable combinations: what if liquidity disappears, correlations jump, and a hedge can no longer be traded?
This requires modelling and financial knowledge, but also imagination and productive disagreement. Which assumption does the team treat as obvious? How could precisely that assumption fail? Good risk work cannot prevent every surprise. It makes clear where the system is vulnerable and which decisions would remain possible.
What these roles share
- Creativity: notice new datasets, models, or solution paths in the first place.
- Probabilistic thinking: work with uncertainty instead of pretending it is gone.
- Scepticism: attack data leaks, spurious correlations, and your favourite ideas.
- Engineering: turn an analysis into a reproducible and reliable process.
- Communication: explain assumptions and risks clearly enough for others to decide.
Where machine learning fits
Machine learning is a toolkit within quantitative finance, not its definition. A linear model may be a better fit for one problem than a neural network. In other cases, gradient boosting, deep learning, or language models help because the volume or form of the data is more complex.
Financial data make ML unusually difficult:
- Markets change, so an old pattern may not remain stable.
- A tiny data leak can make a backtest look convincing and be useless.
- Fees, spreads, and market impact can consume a statistical edge.
- Correlation is not automatically a durable economic explanation.
Baselines, out-of-sample tests, and stress scenarios therefore matter just as much as training the model. Jane Street's description of quantitative research similarly combines experiment design, dataset generation, time-series analysis, feature engineering, and different statistical and ML methods.
A typical quant workflow
- Define the question. What should be estimated, priced, optimised, or hedged?
- Understand the data. Where did they come from, when would they actually have been available, and what biases do they contain?
- Build a simple baseline. Show that a more complex approach beats a clear reference.
- Test the model. Use later test data, realistic costs, and alternative assumptions.
- Check robustness. Does the result hold in other periods and market regimes? Where does it break?
- Make and operate the decision. Set position sizes, limits, monitoring, and the technical implementation.
A beautiful backtest is only an intermediate result. What matters is whether the whole chain from information to controlled decision holds together.
Where do quants work?
This selection is neither a ranking nor exhaustive. It maps a few corners of the industry and links directly to the firms' own career and role pages:
- Market making and quantitative trading: Optiver, IMC, Jane Street, and Hudson River Trading
- Systematic investment management: Citadel, Two Sigma, and D. E. Shaw
- Digital assets and crypto liquidity: Wintermute
Quant roles also exist at banks, asset managers, exchanges, energy traders, and fintech companies. Many of the firms above offer internships and graduate programmes. Compensation is often highly competitive, and entry is correspondingly selective. A single salary number is less informative than three questions: What problems does the team solve, which skills does it test, and would you want to do that work every day?
How can you get started?
You do not need to begin with a finance degree. A practical foundation has four parts:
- probability, statistics, and some linear algebra
- Python and careful work with data
- the basics of markets, financial products, and risk
- small projects whose assumptions and failure modes remain visible
A first project could be a backtest with transaction costs and a genuinely later test period. A Monte Carlo option simulation, a small order-book model, or an investigation of why an attractive signal disappears after realistic costs are equally useful.
Quantitative finance does not mean predicting markets with certainty. It means making uncertainty measurable, testing ideas hard, and building systems that work outside a notebook. That combination of theory, data, and practical implementation is what makes the field so broad.