Alternatives to Keras

Server side

We decided to use Python for our backend because it is one of the industry standard languages for data analysis and machine learning. It also has a lot of support due to its large user base.

• Web Server: We chose Flask because we want to keep our machine learning / data analysis and the web server in the same language. Flask is easy to use and we all have experience with it. Postman will be used for creating and testing APIs due to its convenience.

• Machine Learning: We decided to go with PyTorch for machine learning since it is one of the most popular libraries. It is also known to have an easier learning curve than other popular libraries such as Tensorflow. This is important because our team lacks ML experience and learning the tool as fast as possible would increase productivity.

• Data Analysis: Some common Python libraries will be used to analyze our data. These include NumPy, Pandas , and matplotlib. These tools combined will help us learn the properties and characteristics of our data. Jupyter notebook will be used to help organize the data analysis process, and improve the code readability.

Client side

• UI: We decided to use React for the UI because it helps organize the data and variables of the application into components, making it very convenient to maintain our dashboard. Since React is one of the most popular front end frameworks right now, there will be a lot of support for it as well as a lot of potential new hires that are familiar with the framework. CSS 3 and HTML5 will be used for the basic styling and structure of the web app, as they are the most widely used front end languages.

• State Management: We decided to use Redux to manage the state of the application since it works naturally to React. Our team also already has experience working with Redux which gave it a slight edge over the other state management libraries.

• Data Visualization: We decided to use the React-based library Victory to visualize the data. They have very user friendly documentation on their official website which we find easy to learn from.

Cache

• Caching: We decided between Redis and memcached because they are two of the most popular open-source cache engines. We ultimately decided to use Redis to improve our web app performance mainly due to the extra functionalities it provides such as fine-tuning cache contents and durability.

Database

• Database: We decided to use a NoSQL database over a relational database because of its flexibility from not having a predefined schema. The user behavior analytics has to be flexible since the data we plan to store may change frequently. We decided on MongoDB because it is lightweight and we can easily host the database with MongoDB Atlas . Everyone on our team also has experience working with MongoDB.

Infrastructure

• Deployment: We decided to use Heroku over AWS, Azure, Google Cloud because it is free. Although there are advantages to the other cloud services, Heroku makes the most sense to our team because our primary goal is to build an MVP.

Other Tools

• Communication Slack will be used as the primary source of communication. It provides all the features needed for basic discussions. In terms of more interactive meetings, Zoom will be used for its video calls and screen sharing capabilities.

• Source Control The project will be stored on GitHub and all code changes will be done though pull requests. This will help us keep the codebase clean and make it easy to revert changes when we need to.

Why we built an open source, distributed training framework for TensorFlow , Keras , and PyTorch:

At Uber, we apply deep learning across our business; from self-driving research to trip forecasting and fraud prevention, deep learning enables our engineers and data scientists to create better experiences for our users.

TensorFlow has become a preferred deep learning library at Uber for a variety of reasons. To start, the framework is one of the most widely used open source frameworks for deep learning, which makes it easy to onboard new users. It also combines high performance with an ability to tinker with low-level model details—for instance, we can use both high-level APIs, such as Keras, and implement our own custom operators using NVIDIA’s CUDA toolkit.

Uber has introduced Michelangelo (https://eng.uber.com/michelangelo/), an internal ML-as-a-service platform that democratizes machine learning and makes it easy to build and deploy these systems at scale. In this article, we pull back the curtain on Horovod, an open source component of Michelangelo’s deep learning toolkit which makes it easier to start—and speed up—distributed deep learning projects with TensorFlow:

https://eng.uber.com/horovod/

(Direct GitHub repo: https://github.com/uber/horovod)

Google Analytics is a great tool to analyze your traffic. To debug our software and ask questions, we love to use Postman and Stack Overflow. Google Drive helps our team to share documents. We're able to build our great products through the APIs by Google Maps, CloudFlare, Stripe, PayPal, Twilio, Let's Encrypt, and TensorFlow.

Why we built an open source, distributed training framework for TensorFlow , Keras , and PyTorch:

At Uber, we apply deep learning across our business; from self-driving research to trip forecasting and fraud prevention, deep learning enables our engineers and data scientists to create better experiences for our users.

TensorFlow has become a preferred deep learning library at Uber for a variety of reasons. To start, the framework is one of the most widely used open source frameworks for deep learning, which makes it easy to onboard new users. It also combines high performance with an ability to tinker with low-level model details—for instance, we can use both high-level APIs, such as Keras, and implement our own custom operators using NVIDIA’s CUDA toolkit.

Uber has introduced Michelangelo (https://eng.uber.com/michelangelo/), an internal ML-as-a-service platform that democratizes machine learning and makes it easy to build and deploy these systems at scale. In this article, we pull back the curtain on Horovod, an open source component of Michelangelo’s deep learning toolkit which makes it easier to start—and speed up—distributed deep learning projects with TensorFlow:

https://eng.uber.com/horovod/

(Direct GitHub repo: https://github.com/uber/horovod)

We just launched the Segment Config API (try it out for yourself here) — a set of public REST APIs that enable you to manage your Segment configuration. A public API is only as good as its #documentation. For the API reference doc we are using Postman.

Postman is an “API development environment”. You download the desktop app, and build API requests by URL and payload. Over time you can build up a set of requests and organize them into a “Postman Collection”. You can generalize a collection with “collection variables”. This allows you to parameterize things like username, password and workspace_name so a user can fill their own values in before making an API call. This makes it possible to use Postman for one-off API tasks instead of writing code.

Then you can add Markdown content to the entire collection, a folder of related methods, and/or every API method to explain how the APIs work. You can publish a collection and easily share it with a URL.

This turns Postman from a personal #API utility to full-blown public interactive API documentation. The result is a great looking web page with all the API calls, docs and sample requests and responses in one place. Check out the results here.

Postman’s powers don’t end here. You can automate Postman with “test scripts” and have it periodically run a collection scripts as “monitors”. We now have #QA around all the APIs in public docs to make sure they are always correct

Along the way we tried other techniques for documenting APIs like ReadMe.io or Swagger UI. These required a lot of effort to customize.

Writing and maintaining a Postman collection takes some work, but the resulting documentation site, interactivity and API testing tools are well worth it.

Our whole Node.js backend stack consists of the following tools:

• Lerna as a tool for multi package and multi repository management
• npm as package manager
• NestJS as Node.js framework
• TypeScript as programming language
• ExpressJS as web server
• Swagger UI for visualizing and interacting with the API’s resources
• Postman as a tool for API development
• TypeORM as object relational mapping layer
• JSON Web Token for access token management

The main reason we have chosen Node.js over PHP is related to the following artifacts:

• Made for the web and widely in use: Node.js is a software platform for developing server-side network services. Well-known projects that rely on Node.js include the blogging software Ghost, the project management tool Trello and the operating system WebOS. Node.js requires the JavaScript runtime environment V8, which was specially developed by Google for the popular Chrome browser. This guarantees a very resource-saving architecture, which qualifies Node.js especially for the operation of a web server. Ryan Dahl, the developer of Node.js, released the first stable version on May 27, 2009. He developed Node.js out of dissatisfaction with the possibilities that JavaScript offered at the time. The basic functionality of Node.js has been mapped with JavaScript since the first version, which can be expanded with a large number of different modules. The current package managers (npm or Yarn) for Node.js know more than 1,000,000 of these modules.
• Fast server-side solutions: Node.js adopts the JavaScript "event-loop" to create non-blocking I/O applications that conveniently serve simultaneous events. With the standard available asynchronous processing within JavaScript/TypeScript, highly scalable, server-side solutions can be realized. The efficient use of the CPU and the RAM is maximized and more simultaneous requests can be processed than with conventional multi-thread servers.
• A language along the entire stack: Widely used frameworks such as React or AngularJS or Vue.js, which we prefer, are written in JavaScript/TypeScript. If Node.js is now used on the server side, you can use all the advantages of a uniform script language throughout the entire application development. The same language in the back- and frontend simplifies the maintenance of the application and also the coordination within the development team.
• Flexibility: Node.js sets very few strict dependencies, rules and guidelines and thus grants a high degree of flexibility in application development. There are no strict conventions so that the appropriate architecture, design structures, modules and features can be freely selected for the development.

We just launched the Segment Config API (try it out for yourself here) — a set of public REST APIs that enable you to manage your Segment configuration. A public API is only as good as its #documentation. For the API reference doc we are using Postman.

Postman is an “API development environment”. You download the desktop app, and build API requests by URL and payload. Over time you can build up a set of requests and organize them into a “Postman Collection”. You can generalize a collection with “collection variables”. This allows you to parameterize things like username, password and workspace_name so a user can fill their own values in before making an API call. This makes it possible to use Postman for one-off API tasks instead of writing code.

Then you can add Markdown content to the entire collection, a folder of related methods, and/or every API method to explain how the APIs work. You can publish a collection and easily share it with a URL.

This turns Postman from a personal #API utility to full-blown public interactive API documentation. The result is a great looking web page with all the API calls, docs and sample requests and responses in one place. Check out the results here.

Postman’s powers don’t end here. You can automate Postman with “test scripts” and have it periodically run a collection scripts as “monitors”. We now have #QA around all the APIs in public docs to make sure they are always correct

Along the way we tried other techniques for documenting APIs like ReadMe.io or Swagger UI. These required a lot of effort to customize.

Writing and maintaining a Postman collection takes some work, but the resulting documentation site, interactivity and API testing tools are well worth it.

Our whole Node.js backend stack consists of the following tools:

• Lerna as a tool for multi package and multi repository management
• npm as package manager
• NestJS as Node.js framework
• TypeScript as programming language
• ExpressJS as web server
• Swagger UI for visualizing and interacting with the API’s resources
• Postman as a tool for API development
• TypeORM as object relational mapping layer
• JSON Web Token for access token management

The main reason we have chosen Node.js over PHP is related to the following artifacts:

• Made for the web and widely in use: Node.js is a software platform for developing server-side network services. Well-known projects that rely on Node.js include the blogging software Ghost, the project management tool Trello and the operating system WebOS. Node.js requires the JavaScript runtime environment V8, which was specially developed by Google for the popular Chrome browser. This guarantees a very resource-saving architecture, which qualifies Node.js especially for the operation of a web server. Ryan Dahl, the developer of Node.js, released the first stable version on May 27, 2009. He developed Node.js out of dissatisfaction with the possibilities that JavaScript offered at the time. The basic functionality of Node.js has been mapped with JavaScript since the first version, which can be expanded with a large number of different modules. The current package managers (npm or Yarn) for Node.js know more than 1,000,000 of these modules.
• Fast server-side solutions: Node.js adopts the JavaScript "event-loop" to create non-blocking I/O applications that conveniently serve simultaneous events. With the standard available asynchronous processing within JavaScript/TypeScript, highly scalable, server-side solutions can be realized. The efficient use of the CPU and the RAM is maximized and more simultaneous requests can be processed than with conventional multi-thread servers.
• A language along the entire stack: Widely used frameworks such as React or AngularJS or Vue.js, which we prefer, are written in JavaScript/TypeScript. If Node.js is now used on the server side, you can use all the advantages of a uniform script language throughout the entire application development. The same language in the back- and frontend simplifies the maintenance of the application and also the coordination within the development team.
• Flexibility: Node.js sets very few strict dependencies, rules and guidelines and thus grants a high degree of flexibility in application development. There are no strict conventions so that the appropriate architecture, design structures, modules and features can be freely selected for the development.

Google Analytics is a great tool to analyze your traffic. To debug our software and ask questions, we love to use Postman and Stack Overflow. Google Drive helps our team to share documents. We're able to build our great products through the APIs by Google Maps, CloudFlare, Stripe, PayPal, Twilio, Let's Encrypt, and TensorFlow.

Google Analytics is a great tool to analyze your traffic. To debug our software and ask questions, we love to use Postman and Stack Overflow. Google Drive helps our team to share documents. We're able to build our great products through the APIs by Google Maps, CloudFlare, Stripe, PayPal, Twilio, Let's Encrypt, and TensorFlow.

A huge component of our product relies on gathering public data about locations of interest. Google Places API gives us that ability in the most efficient way. Since we are primarily going to be using as google data as a source of information for our MVP, we might as well start integrating the Google Places API in our system. We have worked with Google Maps in the past and we might take some inspiration from our previous projects onto this one.

We've been using PostgreSQL since the very early days of Zulip, but we actually didn't use it from the beginning. Zulip started out as a MySQL project back in 2012, because we'd heard it was a good choice for a startup with a wide community. However, we found that even though we were using the Django ORM for most of our database access, we spent a lot of time fighting with MySQL. Issues ranged from bad collation defaults, to bad query plans which required a lot of manual query tweaks.

We ended up getting so frustrated that we tried out PostgresQL, and the results were fantastic. We didn't have to do any real customization (just some tuning settings for how big a server we had), and all of our most important queries were faster out of the box. As a result, we were able to delete a bunch of custom queries escaping the ORM that we'd written to make the MySQL query planner happy (because postgres just did the right thing automatically).

And then after that, we've just gotten a ton of value out of postgres. We use its excellent built-in full-text search, which has helped us avoid needing to bring in a tool like Elasticsearch, and we've really enjoyed features like its partial indexes, which saved us a lot of work adding unnecessary extra tables to get good performance for things like our "unread messages" and "starred messages" indexes.

I can't recommend it highly enough.

Often enough I have to explain my way of going about setting up a CI/CD pipeline with multiple deployment platforms. Since I am a bit tired of yapping the same every single time, I've decided to write it up and share with the world this way, and send people to read it instead ;). I will explain it on "live-example" of how the Rome got built, basing that current methodology exists only of readme.md and wishes of good luck (as it usually is ;)).

It always starts with an app, whatever it may be and reading the readmes available while Vagrant and VirtualBox is installing and updating. Following that is the first hurdle to go over - convert all the instruction/scripts into Ansible playbook(s), and only stopping when doing a clear vagrant up or vagrant reload we will have a fully working environment. As our Vagrant environment is now functional, it's time to break it! This is the moment to look for how things can be done better (too rigid/too lose versioning? Sloppy environment setup?) and replace them with the right way to do stuff, one that won't bite us in the backside. This is the point, and the best opportunity, to upcycle the existing way of doing dev environment to produce a proper, production-grade product.

I should probably digress here for a moment and explain why. I firmly believe that the way you deploy production is the same way you should deploy develop, shy of few debugging-friendly setting. This way you avoid the discrepancy between how production work vs how development works, which almost always causes major pains in the back of the neck, and with use of proper tools should mean no more work for the developers. That's why we start with Vagrant as developer boxes should be as easy as vagrant up, but the meat of our product lies in Ansible which will do meat of the work and can be applied to almost anything: AWS, bare metal, docker, LXC, in open net, behind vpn - you name it.

We must also give proper consideration to monitoring and logging hoovering at this point. My generic answer here is to grab Elasticsearch, Kibana, and Logstash. While for different use cases there may be better solutions, this one is well battle-tested, performs reasonably and is very easy to scale both vertically (within some limits) and horizontally. Logstash rules are easy to write and are well supported in maintenance through Ansible, which as I've mentioned earlier, are at the very core of things, and creating triggers/reports and alerts based on Elastic and Kibana is generally a breeze, including some quite complex aggregations.

If we are happy with the state of the Ansible it's time to move on and put all those roles and playbooks to work. Namely, we need something to manage our CI/CD pipelines. For me, the choice is obvious: TeamCity. It's modern, robust and unlike most of the light-weight alternatives, it's transparent. What I mean by that is that it doesn't tell you how to do things, doesn't limit your ways to deploy, or test, or package for that matter. Instead, it provides a developer-friendly and rich playground for your pipelines. You can do most the same with Jenkins, but it has a quite dated look and feel to it, while also missing some key functionality that must be brought in via plugins (like quality REST API which comes built-in with TeamCity). It also comes with all the common-handy plugins like Slack or Apache Maven integration.

The exact flow between CI and CD varies too greatly from one application to another to describe, so I will outline a few rules that guide me in it: 1. Make build steps as small as possible. This way when something breaks, we know exactly where, without needing to dig and root around. 2. All security credentials besides development environment must be sources from individual Vault instances. Keys to those containers should exist only on the CI/CD box and accessible by a few people (the less the better). This is pretty self-explanatory, as anything besides dev may contain sensitive data and, at times, be public-facing. Because of that appropriate security must be present. TeamCity shines in this department with excellent secrets-management. 3. Every part of the build chain shall consume and produce artifacts. If it creates nothing, it likely shouldn't be its own build. This way if any issue shows up with any environment or version, all developer has to do it is grab appropriate artifacts to reproduce the issue locally. 4. Deployment builds should be directly tied to specific Git branches/tags. This enables much easier tracking of what caused an issue, including automated identifying and tagging the author (nothing like automated regression testing!).

Speaking of deployments, I generally try to keep it simple but also with a close eye on the wallet. Because of that, I am more than happy with AWS or another cloud provider, but also constantly peeking at the loads and do we get the value of what we are paying for. Often enough the pattern of use is not constantly erratic, but rather has a firm baseline which could be migrated away from the cloud and into bare metal boxes. That is another part where this approach strongly triumphs over the common Docker and CircleCI setup, where you are very much tied in to use cloud providers and getting out is expensive. Here to embrace bare-metal hosting all you need is a help of some container-based self-hosting software, my personal preference is with Proxmox and LXC. Following that all you must write are ansible scripts to manage hardware of Proxmox, similar way as you do for Amazon EC2 (ansible supports both greatly) and you are good to go. One does not exclude another, quite the opposite, as they can live in great synergy and cut your costs dramatically (the heavier your base load, the bigger the savings) while providing production-grade resiliency.