Josef Teichmann

Josef Teichmann: The Geometer Who Went to Market

His 1999 doctoral thesis was titled "The Theory of Infinite-Dimensional Lie Groups from the Point of View of Functional Analysis." Nothing in that sentence suggests a man who would end up on the programme of every serious quantitative finance conference on earth, or collect the French Academy of Sciences' Louis Bachelier Prize fifteen years later. But an interest rate curve is an infinite-dimensional object, evolving in time, and once Josef Teichmann noticed that, the distance between abstract geometry and the bond market turned out to be very short indeed.

From Lienz to Vienna

Teichmann was born on 27 August 1972 in Lienz, in the Austrian Tyrol. He read mathematics at the University of Graz before moving to the University of Vienna for doctoral work under Peter W. Michor, one of the leading figures in infinite-dimensional differential geometry. The thesis, completed in 1999, sat squarely in that world: Lie groups of infinite dimension, approached through functional analysis rather than through the finite-dimensional intuitions that usually guide the subject.

He then took a post-doctoral position at the Vienna University of Technology and habilitated there in 2002 — the German-speaking academic system's formal licence to hold a chair, and in his case the point at which the trajectory bent.

Infinite Dimensions, Real Money

The move from pure geometry to mathematical finance looks like a career change. It was closer to a change of examples.

Classical option pricing treats a single asset price as a stochastic process and asks what a derivative on it is worth. Interest rates are harder, because what moves is not a number but a curve: the whole term structure of yields across every maturity from overnight to thirty years, wobbling as one connected shape. The Heath–Jarrow–Morton framework, which Teichmann has worked on extensively, models that entire curve as a stochastic process in a function space. The state variable lives in infinitely many dimensions. Which means that the questions a geometer asks — what is the shape of the space, which submanifolds are preserved by the dynamics, when does a flow stay inside a finite-dimensional family — are exactly the questions the finance problem poses.

This is the recurring pattern of his career: bring an unusually strong geometric toolkit to a domain that had been treated mainly with probability and computation, and find structure other people could not see because they were not equipped to look for it.

Rough Paths, Malliavin Calculus, Cubature

His work in stochastic analysis has ranged across the field's most technical machinery. Malliavin calculus — a differential calculus on the space of random paths — supplies sensitivities and smoothness results that are otherwise inaccessible. Rough path theory and the study of stochastic partial differential equations attack the problem of what it even means to solve an equation driven by noise too irregular for classical integration. Affine processes provide the tractable class underlying much of interest rate and volatility modelling.

Alongside the theory sits a persistent concern with computation. Cubature formulas and numerical schemes for stochastic differential equations are not glamorous, but they are how a model becomes a price. Teichmann's willingness to work at both ends — the geometry of infinite-dimensional spaces at one, the numerical scheme at the other — is unusual, and is a large part of why practitioners read him.

Signatures and Randomisation

The most recent turn is towards machine learning, and here too the entry point is structural rather than fashionable. The signature of a path is an infinite sequence of iterated integrals that encodes the path's shape — a canonical set of features arising from rough path theory rather than from an engineer's guesswork. Teichmann's talks on signature theory and mathematical finance pursue the consequence: if signatures give a principled description of a trajectory, they give a principled input to a learning algorithm operating on financial time series.

A parallel line concerns randomisation and reservoir computing, an approach in which large random systems are generated and only a simple readout layer is trained. Its appeal to a mathematician is that its behaviour is analysable in a way that deep networks generally are not. Other recent work concerns the ergodic robust maximisation of asymptotic growth — the question of how to grow capital over the long run when the model itself is uncertain.

Zurich

He joined ETH Zürich as professor in June 2009, in the Department of Mathematics, working within the Stochastic Finance group — the department that has done more than any other in continental Europe to define modern mathematical finance. Since August 2023 he has chaired the department. He lectures and presents worldwide, from Bachelier Colloquiums and SIAM Financial Mathematics meetings to the International Congresses on Industrial and Applied Mathematics.

Recognition arrived early and from several directions: the Prize of the Austrian Mathematical Society in 2005, a START-Preis from the Austrian Science Fund in 2006 — Austria's most prestigious award for young researchers — and the Louis Bachelier Prize of the French Academy of Sciences in 2014.

Why Josef Is Called a Genius

No source in front of me applies the word to him, and it would be dishonest to manufacture one. What the record supports is a specific and uncommon cognitive profile: unusual range, deployed as a transfer mechanism. Most mathematicians who move into finance bring probability. Teichmann brought infinite-dimensional differential geometry, a subject with no obvious connection to markets until you notice that a yield curve is a point in a function space and its evolution is a flow. Doing that once is a good idea. Doing it repeatedly — into rough paths, into affine processes, into signature methods for machine learning — suggests a mind that habitually asks what the geometry of a problem is before asking how to compute it.

The second quality is refusal to specialise downwards. It is normal for a mathematician to become steadily more abstract or steadily more applied. Teichmann has held both, publishing on the structure of infinite-dimensional spaces and on cubature schemes that get implemented, and the Bachelier Prize is precisely the recognition awarded for that combination.

The honest counter-case: he is a working research mathematician of high distinction, not a figure whose name attaches to a theorem that reorganised a field. His prizes are national and disciplinary rather than the Fields Medal or the Abel. Much of his most cited output is collaborative, as is normal in mathematical finance. And the machine learning work is recent enough that its durability is unestablished — signature methods may become standard equipment or a footnote, and it is too early to say which. He belongs among the strongest and most versatile mathematicians in his field, which is a real distinction and a different one from genius.

Legacy

What Teichmann has built in Zurich is a bridge with traffic running both ways: geometry and stochastic analysis feeding into the pricing and hedging problems that banks actually face, and the intractability of those problems generating new pure mathematics in return. Chairing the ETH mathematics department from 2023 puts him in charge of one of the institutions that decides what the next generation of this work looks like. The thesis on infinite-dimensional Lie groups turned out to be less a starting point than a set of tools he has been unpacking, in increasingly unlikely places, ever since.

Achievements

• Louis Bachelier Prize — 2014
• Held posts at TU Wien, ETH Zurich and Pierre and Marie Curie University
• Fields: mathematical finance, stochastic analysis and stochastic process