PyCon Spain 2019

Python is a modern and practical language able to be a strong contender in the new serverless and microservice applications fronts. In this talk, the experience of building a full serverless, microservice application with high computing demands is presented. Although challenging, the application was built using
basic Python modules and a few external libraries. Serverless applications are billed by the number of calls and run time in milliseconds. The run time is a challenge on itself, extra care should be taken initializing modules and the code must be optimized for fast startup. Deployment is also a problem, as Python dependencies can include native compiled code that needs to be deployed to the cloud provider serverless platform. How to share and how to deploy common code is also addressed as an architectural decision. All tools and methods used to improve team productivity and sanity are presented with hints for new projects

Did you know there are multiple ways to raise and capture exceptions? Have you ever wondered if you should raise a built-in exception or create your own hierarchy? Did you ever find it hard to understand what an exception meant?

This talk will go through the decisions needed to raise and capture exceptions when creating a library. We will look at how to translate and handle errors, create your own exceptions, and make exceptions clear and easy to troubleshoot, while also understanding how they actually work, common pitfalls.

PySOA is a general-purpose library created by Eventbrite for writing Python (micro)services, using Redis as a message broker. In this talk we'll learn how it works under the hood.
At Eventbrite most of our services are based on PySOA, serving thousands of requests per second.

Slides: https://docs.google.com/presentation/d/1E_MoEvkJ_gZzGhlS4NjN6juil56PR2fGfWL_J8XTcwo/edit?usp=sharing

Please note the attached presentation in PDF doesn't include animations. For a better experience please use the link to Google Slides above.

Machine learning has become the must-have technology across all industries, largely inspired by successes of new artificial deep neural network applications. Yet, chemical processes have to include process dynamics and control not only to capture and predict the behavior of non-linear systems, but also to optimize their performance. In this talk, an introduction to these concepts using the GEKKO Optimization Suite will be provided.

There are situations when code is not enough. Let me point out some of them:

When someone wants to learn a programming language regardless of the type.
When someone enters to a company with prior knowledge on other languages but not the one is going to work on.
When we want to add a new feature to the application or script.
When a change is require on the behavior of the application.
All these situations have something in common, the code is not enough, it is not enough to get someone to know what the application/script does or even how it does it and that’s when we need more things than just code.
This talk is to shout out the importance of documentation on code.

A data artist (also known as “unicorn”) lives in the intersection of data analysis, decision-making, visualization
and wait for it... ART. They are able, not only to use a number of techniques and tools to transform complex data into delightful and memorable visualizations, but to build their own tools and workflows to create visualizations that go beyond the state of the art. At CARTO we are bringing together data, analysis and visualization, backed by APIs and Python libraries to pave the way for developers, data scientists and data artists in their way to solve geospatial problems.

A spectrum is an essential tool to understand the universe. This is a celestial histogram we can use to disentangle the photons from astronomical bodies. It is an astronomer job to explain the causes behind this distribution of photons. This talk is divided in two sections:  In the first one, we visit a basic theoretical model to explain the incoming light from a galaxy. To do this, we review the tools available in Python to obtain astronomical data and how they can be combined to generate synthetic galaxies. In the second part, we walk the opposite path: We use Python on astronomical data to check the robustness of theoretical models (see Fernández et al. 2019).
The purpose of this presentation is to explain the principles of probabilistic programming. In this case we use PyMC3 (see Salvatier et al 2018) a probabilistic programming library based in Theano (see Theano development team 2016). The author aims to apply a visually simple scientific model to describe some advantages and challenges in neural networks, Bayesian statistics and Monte Carlo samplers.
References:
Fernández, Vital, Elena Terlevich, Angeles I. Díaz, and Roberto Terlevich (2019). “A Bayesian direct method implementation to fit emission line spectra: application to the primordial He abundance determination”. In: Monthly Notices of the Royal Astronomical Society 487.3, pp. 3221–3238. ISSN: 0035-8711. DOI:10.1093/mnras/stz1433
Salvatier, John, Thomas V. Wiecki, and Christopher Fonnesbeck (2016). “Probabilistic programming in Python using PyMC3”. In: PeerJ Computer Science 2, e55. ISSN: 2376-5992. DOI: 10.7717/peerj-cs.55.
The Theano Development team et al. (2016). “Theano: A Python framework for fast computation of mathematical expressions”. In: arXiv:1605.02688 [cs]. arXiv: 1605.02688.

[START] I spend the main part of my life in front of my screen, coding andtalking to my computer what to do thanks to my terminal. Command lines is part of my everyday life. [THEN] One day I started to write my own command lines. [SO] In this talk I’d like to share with you how you can easily create command lines in python. I hope it would be helpful. First, we will see how with setuptools and PyPI, we can transform a simple python script onto a crossplatform command line. Then we will see the POSIX format. Finally, We’ll see how with docopt we can define the interface for your command line, and automatically generate a parser for it.

Machine learning unfolds its transformational power when applied at scale in real-world environments. e-Commerce and other retail tech organizations have to make day by day non-trivial business decisions, many of which have typically been done with simple heuristics or other types of codified business rules. Machine learning models have proven to outperform these heuristics in automating complex decision making, given enough training data.

For a business to be able to transform itself from a data-collecting organization to a data-driven one, artificial intelligence algorithms, complex data processing pipelines and an accordingly designed IT architecture must be in place. The value of such applied machine learning systems that are deeply integrated in an organization’s software stack is more than the sum of its individual parts.

We will show the nuts and bolts of an end-to-end model training, deployment and scoring system that serves model decisions for underwriting e-Commerce return transactions in real-time, with a 100 % cloud-based infrastructure. Our platform includes a unified model building pipeline that leverages the Python ecosystem: custom and fully integrated preprocessing and feature engineering using scikit-learn, Bayesian hyperparameter optimization of the whole learning process using hyperopt, experiment tracking with MLflow and a REST API for serving predictions in production with Flask, among others.

Having a team of two interrogators who adopt different strategies to a subject matter is a good cop/bad cop approach to solving some criminal cases. Here, our case is not criminal. We want to solve the problem of selling products to people bearing in mind that the users choice is subjective.

There has been a major disruption in how advertising is being done recently. The typical radio jingle, newspaper adverts and television commercials are gradually becoming redundant. Brands want their products displayed on webpages, mobile applications and other intelligent platforms. This diversion from the more conventional forms of marketing has led to major successes in sales as well as brand loyalty and recognition. On the other hand, users get spammed and sometimes even irritated by the volume of display advertisements exposed to them.

What then is the solution? Users would be more interested in products (good or bad) if they met their needs. In this talk, we would run a simple A/B test where different variations of products are tested and extremely small improvements are being estimated. We would also take it a step further by discussing the basics of a click prediction engine and lookalike modeling.

The presentation will provide a describe a test case of the application of Python to the extraction, manipulation data analysis of Force Curve Spectroscopy in Atomic Force Microscopy. It will demonstrate how to parse text files, plot relavent properties, sort through extraneous data, and interface with large amounts of force curve data.

GDPR came into effect in 2018 and security become even more important. Whats more, it became a shared responsibility and not everybody realizes this. Security incidents and problems continue to happen and I would like to share my experiences as a Security Engineer working closely with developers how to tackle the cooperation with developers and what tools and techniques one can use to spot security problems before they will be misused.

Aunque difíciles de identificar, los patrones de diseño son una herramienta clave en la construcción de cualquier sistema software y nos ayudan a plantear soluciones que son integrables rápidamente, con poco
esfuerzo, cuya conceptualización ha sido ya validada por múltiples desarrolladores y estableciendo un enfoque que puede ser entendido fácilmente, sin importar el lenguaje de programación o tecnología
subyacente.

En portales que orquestan múltiples servicios de distintas naturalezas se hacen incluso más valiosos, ya que reducen el tiempo para la integración de nuevos desarrollos facilitando la conceptualización de la solución a
implementar.

Durante la charla daremos un repaso a los patrones de diseño que nos han ayudado a aterrizar los problemas a los que nos enfrentamos día a día en la construcción de una plataforma fintech que gestiona servicios de inversión para distintos países.

La charla cubre desde patrones incorporados por Python y Django a patrones implementados de forma específica para dar solución a nuestra problemática, como el patrón de diseño Adaptador para establecer
interfaces flexibles con third parties o el patrón Pipeline para el procesamiento de procesos complejos en etapas.

Companies fulfill the needs of the population through the goods and services than they produce. Therefore, companies impact society by selling products to individuals. To quantify this impact we can analyze people’s consumption. Although traditionally this consumption is measured through surveys, we took advantage of modern Data Science techniques to analyze big data sources of real consumption data.

In this talk I will explore some example of the economic and data science analysis that we developed in Python to understand people’s consumption. I will explain how we are able to apply Angus Deaton Elasticity Theory to Big Data showing which goods are first-necessity-goods for the Mexican population. I will introduce Manifold Learning techniques to cluster individuals into meaningful consumption groups that transcends demographic features. Finally, I will show how to analyze consumption data as a complex system to compute the real Maslow’s hierarchy of needs based on the sophistication of individual consumers.