Avatar

Salvatore Mercuri
Knowledge Transfer Associate

School of Mathematics, The University of Edinburgh

Email: smercuri (at) ed.ac.uk

Address:
School of Mathematics, University of Edinburgh
James Clerk Maxwell Building
The King's Buildings
Peter Guthrie Tait Road Edinburgh
EH9 3FD, United Kingdom

Office: 5422





I am currently a Knowledge Transfer Associate in Digital Mathematics Assessment in the School of Mathematics at the University of Edinburgh, working in partnership with IDEMS International CIC.

My current primary research area encompasses digitisation of mathematics research and education, with a focus in developing the STACK computer-aided assessment system to automate assessment of proof questions as well as to broaden applications to machine learning and data science. Within this area I am also interested in formal proof verification for mathematical research and education, in particular with Lean theorem prover. My mathematical research background is in the intersection of algebraic and analytic number theory, with research focusing on the algebraicity of metaplectic automorphic forms and the broader context of the Langlands program. Since then I have been a researcher in machine learning in scientific and financial industries, leading to my interests circling back to the application of computational and statistical methods to enhance mathematics research and education.


Employment

Knowledge Transfer Associate at the University of Edinburgh and IDEMS International
2023 - 2025

Research Data Scientist at NatWest Group
2021 - 2023

Mathematical Researcher at ThinkTank Maths
2020 - 2021

Education

PhD in Mathematics
2015 - 2019

Department of Mathematics, Durham University
Advisor: Thanasis Bouganis

Thesis: The arithmetic of metaplectic modular forms


MSc. in Pure Mathematics
2014 - 2015

Imperial College London
Advisor: Kevin Buzzard


BSc. in Mathematics
2010 - 2014

University of Edinburgh


Publications

  1. (with D. Batra, R. Khraishi) Conformal Predictions for Longitudinal Data, preprint (2023).
  2. (with G. Cowan, R. Khraishi) Modelling customer lifetime value in the retail banking industry, preprint (2023).
  3. (with R. Khraishi, R. Okhrati, D. Batra, C. Hamill, T. Ghasempour, A. Nowlan) An Introduction to Machine Unlearning, (2022).
  4. Algebraicity of metaplectic L-functions, (arXiv link), J. Number Theory 219 (2021), 109-161.
  5. (with T. Bouganis) On the Rankin–Selberg method for vector-valued Siegel modular forms, (arXiv link), Int. J. Number Theory 17 (2020), no. 5, 1207-1242.
  6. p-adic L-functions on metaplectic groups (arXiv link), J. London Math. Soc., 102 (2020), 229-256.
  7. The p-adic L-function for half-integral weight modular forms, manuscripta math., 161 (2020), 61-91.

Talks


Conferences and workshops attended


Resources

STACK

STACK is open-source software that is used for automated digital assessment of mathematics within universities and schools globally. To get get started writing questions, install STACK and then use the author quick-start guide.

There are increasing ways to assess proof questions in STACK and my work so far has helped enable assessing proofs as drag-and-drop reordering questions through a parsons block. This was released in Dec. 2023 in STACK v4.5.0. See the documentation for how to write and assess these types of questions in STACK.


Lean

The Lean theorem prover is an automated theorem prover that has grown rapidly as a tool for research and for education in both mathematics and AI. Mathlib is an open-source library for Lean containing a large and growing body of mathematical theory in Lean.

A great place to start learning is the Natural Number Game. After that, a good next step would be to install Lean and follow Imperial College London's "Formalising Mathematics 2024" course on GitHub, as well as reading Mathematics in Lean.

I have a GitHub repository formalising the proof of the local compactness of the adele ring of a number field here.


FLINT

The images below and on the home page were made using the C library Arb, which is now part of FLINT (Fast Library for Number Theory). They show the complex plots of modular forms, including the half-integral weight Dedekind eta function, and other elliptic functions as well as L-functions. The images use colour to plot the phase of the output complex number of the function for each complex input, being forgetful of magnitude.

As well as facilitating the production of high resolution images of intricate complex objects through the use of ball arithmetic, FLINT has been used for computational research purposes such as the highest-known (as of January 2024) verification of the Riemann Hypothesis, Dave Platt & Tim Trudgian (2020).




Outreach

Member of the STACK Professionals Network, University of Edinburgh/IDEMS International
2023 -