I am a Data Scientist Director (Principal Data Scientist) at Nike. Previously, I was a Data Scientist working for Celect, Inc (acquired by Nike), and I was a postdoctoral researcher at the Center for Theoretical Physics at MIT. I got my bachelor's degree, master's degree and Ph.D. in Theoretical Physics at the Institute for Theoretical Physics at ETH Zurich.

I am working on personalization, forecasting, distributed computation and mathematical optimization.

Papers and preprints

Cyril Stark, Expressive Recommender Systems through Normalized Nonnegative Models, AAAI-16 Conference, Phoenix AZ (February 2016); arXiv:1511.04775

Cyril Stark and Aram Harrow, Compressibility of positive semidefinite factorizations and quantum models, accepted for publication in IEEE Transactions on Information Theory in 2016, accepted for ISIT 2015, arXiv:1412.7437

Cyril Stark, Recommender systems inspired by the structure of quantum theory, arXiv:1601.06035 (January 2016).

Cyril Stark, Amortized inference through normalized nonnegative models, Black Box Learning and Inference Workshop, NIPS, Montreal QC (December 2015).

Cyril Stark, Top-N recommendations from expressive recommender systems, arXiv:1511.06718 (November 2015).

Cyril Stark, Learning optimal quantum models is NP-hard, arXiv:1510.02800 (October 2015); Physical Review A 97.2 (2018): 020103.

Cyril Stark, Global Completability with Applications to Self-Consistent Quantum Tomography, Communications in Mathematical Physics, 348: 1. doi:10.1007/s00220-016-2760-2 (2016).

Cyril Stark, Self-consistent tomography of the state-measurement Gram matrix, Phys. Rev. A 89(5):052109 (May 2014).

Cyril Stark, Simultaneous Estimation of Dimension, States and Measurements: Computation of representative density matrices and POVMs, arXiv:1210.1105 (October 2012).

Cyril Stark, Lode Pollet, Atac Imamoglu and Renato Renner, Localization of Toric Code Defects, Phys. Rev. Lett. 107:030504 (July 2011).

Selected Talks

Physics-inspired techniques for association rule mining, Quantum Machine Learning conference, Perimeter Institute for Theoretical Physics, Waterloo, ON (August 2016; invited)

Identifiability of matrix factorizations, Geometry in Machine Learning Workshop, International Conference on Machine Learning, New York NY (June 2016)

Global completability of matrix factorizations, First International Workshop on Geometry and Machine Learning, Annual Symposium on the Theory of Computing/International Symposium on Computational Geometry, Boston MA (June 2016)

Expressive Recommender Systems through Normalized Nonnegative Models, AAAI-16 Conference, Phoenix AZ (February 2016)

Learning quantum models, Harvard University, Cambridge, MA (January 2016).

Quantum models for non-physical data at the example of item recommendation, Quantum Machine Learning Workshop, Conference on Neural Information Processing Systems (NIPS), Montreal QC (December 2015)

Amortized inference through normalized nonnegative models, Black Box Learning and Inference Workshop, NIPS, Montreal QC (December 2015)

Learning quantum models for physical and non-physical data, Perimeter Institute for Theoretical Physics, Waterloo, ON (October 2015; invited).

Compressibility of positive semidefinite factorizations, 22nd International Symposium on Mathematical Programming, Pittsburgh PA (July 2015).

Inference of low-dimensional quantum models is NP-hard, University of Bristol, Bristol UK (June 2015).

Compressibility of positive semidefinite factorizations and quantum models, IEEE International Symposium on Information Theory, Hong Kong (June 2015).

Inference of low-dimensional quantum models is NP-hard, ETH Zurich, Zurich, Switzerland (June 2015).

Robustness of the Hilbert space dimension under measurement uncertainty, MIT, Cambridge MA (February 2015).

Compressibility of positive semidefinite factorizations and quantum models, Coogee, Sidney (January 2015; invited).

Compressibility of positive semidefinite factorizations and quantum models, National University of Singapore, Singapore (January 2015).