About

I am a Data Science Initiative Postdoctoral Fellow at Harvard University, working in the lab of Pierre Jacob. Previously, I did my PhD in Statistics at Columbia University with Liam Paninski. During my PhD days I had a stint at Google Brain as an intern, where I was hosted by Jasper Snoek.

The picture above is from my hometown: Santiago, Chile. There I obtained by B.S. in Mathematical Engineering at Universidad de Chile.

Interests

I am a computational statistician who is interested in developing scalable, robust and interpretable methods and algorithms for making sense of the large volumes of data that are routinely collected with modern technologies in science. As of today my focus is mostly methodological; some questions I am interested in are (the list is not exhaustive): what makes popular algorithms work well in certain regimes (e.g. dimension) but not in others, what makes models robust, and how can we desing robust algorithms. I believe the three statistical, computational and mathematical perspectives are quite relevant to tackle those questions.

I do enjoy working with real data: in my PhD I developed algorithms for extracting neural signals from recordings, and I am looking forward to starting new collaborations involving data.

News

• Sept 01st, 2018. Started postdoc!

• Feb 14th, 2018. I gave the talk Statistical Machine Learning Methods for Large-scale Neural Data Analysis at the Vector Institute, Toronto.

• Jan 29th, 2018. Our paper Learning latent permutations with Gumbel-Sinkhorn networks (joint with D. Belanger, S. Linderman and J. Snoek) was accepted to ICLR 2018.

• Jan 22nd, 2018. I gave the talk Optimal Transport and applications to Data Science at the Center for Mathematical Modeling, Universidad de Chile.

• Dec 22nd, 2017. Our paper Reparameterizing the Birkhoff Polytope for Variational Permutation Inference (joint with S. Linderman, H. Cooper, J. Cunningham and L. Paninski) was accepted to AISTATS, 2018. Nos vemos en Lanzarote!

• Dec 1st, 2017. I will be presenting two workshop papers at NIPS: Our work on Sinkhorn Networks will be presented in the Optimal Transport & Machine Learning Workshop. Also, our work on permutation inference for C. elegans data will be presented in the Worms Information Processing Workshop.

• Nov 13rd, 2017. Our methods paper Electrical stimulus artifact cancellation and neural spike detection on large multi-electrode arrays is now in press at PLOS Computational Biology. We think this is a significant step towards enabling closed-loop experiments featuring high-resolution neural electrical stimulation.