Curriculum Vitae

E-mail: mail@tsmithe.net

Born: 8th February 1992, London, UK

Nationality: British citizen

Languages: English, French

Education

MSc Complex Systems Science (Erasmus Mundus)

Chalmers University (year 1), École Polytechnique (year 2), Warwick University

  • Dates: <2014-09-01 Mon>–<2016-07-31 Sun>
  • Year 2 research project

    Group for Neural Theory, Laboratoire de Neurosciences Cognitives, École Normale Supérieure, Paris

    For my Master's thesis, I will be working with Sophie Deneve, looking at the balance between excitation and inhibition in spiking neural networks, and the oscillatory effects of delayed inhibitory responses, with a view to understanding how these properties affect the statistics of the neural code, and whether it is possible to explain oscillatory effects observed in cortex and behaviour with respect to attention and schizophrenia.

  • Year 2 modules; in the CogMaster programme at the École Normale Supérieure (rue d'Ulm, Paris) and the mathematics / vision / learning (MVA) masters at the École Normale Supérieure (Cachan):
    • Theoretical neuroscience (CogMaster), including mathematical models of: neurons, synapses, coding, learning, memory, vision, and navigation.
    • Modelling of action, decision and volition (CogMaster), including: executive function, cognitive/attentional/motivational control, perceptual decision-making, reinforcement learning, and neuroeconomics.
    • Perception, action and attention (CogMaster), including: sensorimotor coding and decision-making, brain coding of movement, visual stability, stimulus-driven and voluntary action control, ideomotor theory.
    • Probabilistic graphical models (MVA), including: linear and logistic regression, K-means and EM algorithms, Gaussian mixtures, directed and undirected graphical models, sum-product algorithm, information theory, exponential families, factor analysis, hidden Markov models, approximate and variational inference, sampling, Bayesian methods, and model selection.
  • Year 1 research project in polymer physics; article submitted to Physical Review E

    Department of Physics, University of Gothenburg

    • Under the supervision of Bernhard Mehlig, I computed the finite-size corrections for an asymptotic theory of the statistics of confined polymers in the extended de Gennes regime using high-performance Monte Carlo simulations implementing the pruned-enriched Rosenbluth method.
    • I performed an experimental analysis of microscopy videos of RecA-coated DNA confined in the Odijk regime and thereby elucidated the nucleation time for the formation of hairpin bends.

BSc (Hons) Open (Mathematics)

Open University

BA (Hons) Psychology, Philosophy and Physiology

Magdalen College, University of Oxford

  • Degree class: Upper Second
  • Dates: <2010-10-01 Fri>–<2013-06-30 Sun>
  • Finals papers
    • Neural and Computational Mechanisms of Decision-Making under Uncertainty [1st class]
    • Practical classes in neural networks [1st]
    • Computational Neuroscience [2.1]
    • Cognitive Neuroscience [1st]
    • Perception [1st]
    • Memory, Attention and Information Processing [2.1]
    • Language and Cognition [2.2]
    • History of Philosophy from Descartes to Kant [1st]
    • Philosophy of Mind [2.1]
    • Philosophy of Cognitive Science [2.1]
    • Experimental Design and Statistics [2.1]
    • Formal Logic [2.2]

The Judd School, Tonbridge, Kent

  • Winner: 2010 Lloyd-Morgan Prize for All-round Academic Excellence
  • 9 A* grades, 1 A grade at GCSE
  • 2 A* grades, 3 A grades at A-level
    • Mathematics [ A* ]
    • English Literature [ A* ]
    • French [ A ]
    • Music [ A ]
    • Chemistry [ A ]
  • 1 A grade at AS-level (supplementary to the above A-levels)
    • Further Mathematics [ A ]

Conference contribution

  1. T. St Clere Smithe: PyViennaCL – Very easy GPGPU linear algebra; Talk: SciPy 2014, Austin, Texas.

Software

  • Programming languages: C, C++, Haskell, Mathematica, MATLAB, Python, Shell.

PyViennaCL

Institute for Analysis and Scientific Computing, Vienna University of Technology

During the Google Summers of Code 2013 and 2014, under the aegis of the Institute for Analysis and Scientific Computing at TU Wien, I proposed, designed and developed PyViennaCL, the Python-language bindings for the open source ViennaCL library for high-performance GPGPU linear algebra and scientific computing, which is written in OpenCL, CUDA and C++. This involved designing a language-agnostic abstract representation for algebraic expressions, which is then passed to a kernel generator and scheduler, thereby enabling scientific code written in Python seamlessly to produce very high performance on a variety of computing architectures, from single-core machines to large heterogeneous systems. I also wrote substantial documentation and a test suite to guarantee correctness.

ROOT and cppyy

CERN

During the Google Summer of Code 2015, I worked with the "software for experiments" group at CERN on cppyy. cppyy uses the new cling C++ interpreter (built on top of clang and LLVM) in order to provide JIT-compiled runtime access to arbitrary C++ code from Python. My work aims to simplify the API, and provide new features for the manual tuning of the generated Python interface. Ultimately, this project will greatly improve the writing of high-performance scientific code in Python, by eliminating the cumbersome work of writing bindings for C++ code, and thereby allowing developers to drop easily into C++ for performance-sensitive routines.