I'm a Machine Learning Engineer working at Landing AI, focusing primarily on the LandingLens platform (backend, infra, etc.). Before that I was at Falabella Retail, one of the biggest retails in Latin America, leading one of the first data science initiatives that involved implementing and deploying product ranking algorithms for the website. My first years after graduating I worked primarily in academia. I was a researcher for ALGES Laboratory at Universidad the Chile, working on bayesian machine learning algorithms for the mining exploration process. During that time I co-founded Universidad the Chile's official Machine Learning group and also published a best paper award at DSP 2017.

I love learning in general, and would describe myself as someone who does great at things that spark my interest and suck at ones that don't (looking at you, law and geography). My goal is to help as much as I can bridging the gap between Machine Learning and Software Development, and have a particular interest in MLOPs even before the term existed.

Timeline

Check my resume or my LinkedIn for more details of what I've been up to lately!

Public work

Papers, presentations, teaching material, etc.

Initialising Kernel Adaptive Filters with Probabilistic Inference

I. Castro, C.Silva, F. Tobar / IEEE DSP 2017

We present a probabilistic framework for both (i) determining the initial settings of kernel adaptive filters (KAFs) and (ii) constructing fully-adaptive KAFs whereby in addition to weights and dictionaries, kernel parameters are learnt sequentially.

Rapid Multivariate Resource Assessment

F. Soto, M. Garrido, G. Díaz, C. Silva / GEOMIN-MINEPLANNING 2017

We developed a tool integrating different visualizations and algorithms in order to make the geological resource assessment process faster. We focused mainly in three aspects: a software responsive in terms of usability, a software with high performance capabilities in terms of parallel and distributed execution of geostatistical algorithms, and a software that allows to re-execute the same resource assessment task in different time scales with little overhead for the user.

Modelamiento Semántico del Entorno de un Robot utilizando información RGB-D

C. Silva / DIE, Universidad de Chile, 2016

The main contribution of this work is a review of Probabilistic Graphical Model literature and its applications on Computer Vision tasks, specifically Semantic Segmentation. In particular, this work serves as a guide on how to model images as undirected graphs and use a general purpose solver for optimizing the outputs of a classification model without additional data.

A Summary of Capsule Networks with an Application on FashionMNIST

C. Silva / EVIC 2017

This work is an application of a novel approach to how neuron outputs in neural networks are represented. These neurons are called capsules and they output vectors instead of scalars. Using these units with a dynamic routing mechanism between layers, a clothes dataset was trained and evaluated to verify the properties of this new approach.

Projects

Things I've done for fun.

KAFTools

A Python library with common Kernel Adaptive Filtering variants.

ML and DS Notebooks

Collection of personal Machine Learning notes and projects.

Dockerfiles

Collection of dockerfiles I created that can be useful for others.

Misc.

Stuff I think it's worth mentioning :)

Random facts

• I spend an unusual amount of time playing videogames. Always thought tech would be full of gamers but so far I've barely met people that have consoles 💔.
• I'm from Punta Arenas, Chile. It's one of the southernmost cities in the world. Great place for a quiet life, not sure if good for people who like computers and internet 🙃.
• I once had a Youtube channel where I played videogames with a friend just for fun. Wanna see? It's in spanish though 🕹.
• I got a GCP Data Engineer certification just because someone told me it was impossible to get it in less than a month (I won the bet, lol). Never renewed because a month after that I moved to a job that uses AWS ¯\_(ツ)_/¯.

Brag section

• Won an internal hackathon in 2020 during my time at Landing AI. We set up an inference server that can automatically label data when users are working with the LandingLens Labeling Tool, effectively helping labelers to label faster.
• Won an internal award in 2018 during my time at Falabella due to the outstanding results of our product ranking algorithm. This was not personalization in case you're thinking that, but rather a smarter way of listing products in categories for the general website user.
• Won a best paper award at the DSP'2017 conference held at Imperial College with Iván Castro and Felipe Tobar. Here is the link to the paper.
• Co-founded the official Machine Learning Community at Universidad de Chile in 2016, Beauchef Machine Learning. The group has over 500 members as of today and has served as a hub to promote talks, job offers, and more.
• My undergraduate thesis has over 2,000 reads at ResearchGate. The reason eludes me (maybe because it's in spanish?) but I've used it multiple times as a reference to help people understand the very basics of Markov Random Fields for Computer Vision.