Software Engineering Daily

Data labeling is a major bottleneck in training and deploying machine learning and especially NLP. But new tools for training models with humans in the loop can drastically reduce how much data is required. Humanloop is a platform for annotating text and training NLP models with much less labelled data. Raza Habib, founder of Humanloop, joins the show to to talk about NLP workflows and his work on Humanloop. Sponsorship inquiries: [email protected]

A customer data platform such as Segment allows developers to build analytics and workflows around customer data such as purchases, clicks, and other interactions. These customer data platforms (CDP) are often tightly coupled to an underlying data warehouse technology. Hightouch is a platform that provides an unbundled CDP--a platform that sits on top of your own data warehouse. The Hightouch team joins the show to talk about what they are building and the CDP ecosystem as a whole. Sponsorship inquiries: [email protected]

Fivetran is a company that builds data integration infrastructure. If your company is performing ELT or ETL jobs to move data from one place to another, Fivetran can help with that movement from source to destination. Once the data is moved into a data warehouse, a tool called DBT (data build tool) can be used to transform the data more effectively. We have done shows previously about Fivetran and DBT. In today’s episode, George Fraser of Fivetran returns to discuss the cross section of these two technologies, and what his company is doing around that integration point. Sponsorship inquiries: [email protected]

Staff engineer is a job title that suggests the engineer has deep expertise, and considerable experience. More and more companies are adopting a “staff engineer track” where an engineer can work to become a staff engineer. What is the role of staff engineer? Is it a management role or an individual contributor? What are the expectations and obligations of staff engineer? Will Larson is an experienced engineer who has worked at Stripe and other prominent tech companies. He joins the show to talk about the role of staff engineering, and the material he has written about it. Sponsorship inquiries: [email protected]

The Salesforce Ecosystem has thousands of developers, designers, product people, and entrepreneurs engaging with each other. Salesforce exposes APIs and SDKs that allow people to build infrastructure on top of the Salesforce platform. In a previous episode, we explored how the ecosystem works as a whole. In today’s show, Chuck Liddell joins the show to talk about how developers themselves engage with Salesforce. Chuck is CEO of Valence, a Salesforce AppExchange ISV that adds native integration middleware to Salesforce. Sponsorship inquiries: [email protected]

Pair programming allows developers to partner on solving problems and learn from each other more effectively. Pair programming has become harder to do as remote work has become more prevalent. GitDuck is a tool to enable more effective pair programming. Dragos Fotescu and Thiago Monteiro are the founders of GitDuck, and they join the show to explain what they have built and their motivation behind it. Sponsorship inquiries: [email protected]

Effective data science requires clean data. As data moves through the data pipeline, there may be errors introduced. Errors can also arise from code changes, database migrations, and other forms of data movement. How can you ensure data quality within a fast moving, dynamic data system? Datafold is a company built around data quality management. It allows users to compare tables and databases, as well as automate data QA. Gleb Mezhanskiy is a founder of Datafold and joins the show to talk about the data quality space and what he is building with Datafold. Sponsorship inquiries: [email protected]

Federated learning is machine learning without a centralized data source. Federated Learning enables mobile phones or edge servers to collaboratively learn a shared prediction model while keeping all the training data on device. Mike Lee Williams is an expert in federated learning, and he joins the show to give an overview of the subject and share his thoughts on its applications. Sponsorship inquiries: [email protected]

For all the advances in software development over the years, one area that has seen minimal improvement is the terminal. Typing commands into a black text interface seems antiquated compared to the dynamic, flashy interfaces available in web browsers and modern desktop applications. Fig is a visual terminal assistant with the goal of changing that. Fig sits next to the developer’s normal terminal and enhances the terminal experience. The founders of Fig, Brendan Falk and Matt Schrage, join the show today to discuss how Fig works and why it is useful to have an enhanced terminal. Sponsorship inquiries: [email protected]

Cloud resources can get out of control if proper management constraints are not put in place. Cloud Custodian enables users to be well managed in the cloud. It is a YAML DSL that allows you to easily define rules to enable a well-managed cloud infrastructure giving security and cost optimization. Kapil Thangavelu works on Cloud Custodian and he joins the show to talk about modern cloud management and what he is building with Cloud Custodian. Sponsorship inquiries: [email protected]

Predicting the spread of COVID-19 is not easy. The best methods we have available require us to extrapolate trends from a large volume of data, and this requires the construction of large-scale models. Because of the expertise needed for developing these models, Silicon Valley engineers were brought in to help develop a maintainable model. Two of these engineers are Josh Wills and Sam Shah, and they join the show to talk about the engineering behind the COVID model, and their work to build it. Sponsorship inquiries: [email protected]

Machine learning models require training data, and training data needs to be labeled. Raw images and text can be labeled using a training data platform like Labelbox. Labelbox is a system of labeling tools that enables a human workforce to create data that is ready to be consumed by machine learning training algorithms. The Labelbox team joins the show today to discuss training data and how to label it. Sponsorship inquiries: [email protected]

Containers and virtual machines are two ways of running virtualized infrastructure. Containers use less resources than VMs, and typically use the runc open source container runtime. Sysbox is a containerization runtime that offers an alternative to runc, and allows for the deployment of Docker or Kubernetes within a container. Cesar Talledo is the founder of Nestybox, a company built around the Sysbox runtime. He joins the show to talk about container runtimes and his new company. Sponsorship inquiries: [email protected]

Firebase is well-known as a platform that makes it easy to build real-time applications quickly and easily. Firebase was acquired by Google, and has been turned into a large platform that runs on top of Google Cloud. Firebase is closed-source, which leads to a different ecosystem than open source platforms. Supabase is a new open source alternative to Firebase, built on Postgres and Elixir. Paul Copplestone is the founder of Supabase and he joins the show to talk through what he is building. Sponsorship inquiries: [email protected]

Development environments are brittle and hard to manage. They lack the kind of fungibility afforded by infrastructure-as-code. Gitpod is a company that allows developers to describe development environments as code to make them easier to work with, and enabling a more streamlined GitOps workflow. Johannes Landgraf and Sven Efftinge are creators of Gitpod and they join the show to discuss the product and the motivation for building it. Sponsorship inquiries: [email protected]

Training a computer vision model is not easy. Bottlenecks in the development process make it even harder. Ad hoc code, inconsistent data sets, and other workflow issues hamper the ability to streamline models. Roboflow is a company built to simplify and streamline these model training workflows. Brad Dwyer is a founder of Roboflow and joins the show to talk about model development and his company. Sponsorship inquiries: [email protected]

Databases are the source of truth for every company. Editing the data in the database normally requires writing a query in SQL or a domain specific querying language–languages that are only accessible to engineers and highly technical people. BaseDash is a tool for interfacing with a database without requiring the usage of a query language. It allows the user to interface with the database as easily as a spreadsheet. Max Musing is a founder of BaseDash, and he joins the show to talk about how it works and why he built it. Sponsorship inquiries: [email protected]

Machine learning models are only as good as the datasets they’re trained on. Aquarium is a system that helps machine learning teams make better models by improving their dataset quality. Model improvement is often made by curating high quality datasets, and Aquarium helps make that a reality. Peter Gao works on Aquarium, and he joins the show to talk through modern machine learning and the role of Aquarium. Sponsorship inquiries: [email protected]

Ray is a general purpose distributed computing framework. Ray is used for reinforcement learning and other compute intensive tasks. It was developed at the Berkeley RISELab, a research and development lab with an emphasis on practical applications. Ion Stoica is a professor at Berkeley, and he joins the show to talk about the present and future of the Ray framework. Sponsorship inquiries: [email protected]

A private network connects servers, computers, and cloud instances. These networked objects are often separated by firewalls and subnets that create latency and complication. David Crawshaw is the CTO of Tailscale, a company that works to make private networks easier to build and simpler to configure and maintain. David joins the show to talk about private networks and the implementation of Tailscale. Sponsorship inquiries: [email protected]

Pachyderm is a system for data version control. Code has been version controlled for many years, but not data. In previous episodes with Joe Doliner, we explored the evolution of Pachyderm. In today’s show, we talk about the state of the company in 2020, as well as Pachyderm Hub, and end-to-end machine learning and data lineage product. Sponsorship inquiries: [email protected]

Deno is a runtime for JavaScript applications. Deno is written in Rust, which changes the security properties of it. Parts of Deno are also written in TypeScript, which are causing problems in the compilation and organization of Deno. Elio Rivero is an engineer who has studied Deno and TypeScript, and he joins the show to talk about the newer JavaScript runtime and the issues caused by TypeScript. Sponsorship inquiries: [email protected]

Developer tooling and infrastructure is a fruitful area for investing. A wide variety of technologies can have large investment outcomes based on the fact that there are lots of engineers and businesses are willing to pay for products that give those engineers a higher degree of leverage. Lee Edwards is a partner with Root Ventures. His focus is on hard problems within software, and he joins the show to talk about the thesis of his firm, as well as his personal beliefs on what makes a good investment. Sponsorship inquiries: [email protected]

Salesforce is a platform with a large number of developers, ISVs, and companies built on top of it. There is a thriving ecosystem of applications built and managed around Salesforce, leading to an important set of relationships and integration points between Salesforce and the other entities involved with the company. Kevin Poorman works at Salesforce as a developer evangelist, helping to strengthen the relationships in the Salesforce ecosystem. Kevin joins the show to talk about Salesforce and the applications that connect to it. Sponsorship inquiries: [email protected]

Twitter is a social media platform with billions of objects: people, tweets, words, events, and other entities. The high volume of information that gets created on Twitter everyday leads to a complex engineering problem for the developers building the Twitter search index. Nico Tonozzi is an engineer at Twitter. He joins the show to talk through the problem space of search at Twitter, as well as some recent challenges that he had to tackle in the continuously changing Twitter product. Sponsorship inquiries: [email protected]

Robinhood is a platform for buying and selling stocks and cryptocurrencies. Robinhood is complex, fast-moving, and financial, and together these things require high quality engineering in distributed systems, observability, and data infrastructure. Jaren Glover is an engineer at Robinhood, and he joins the show to talk about the problem space within Robinhood, as well as the specific DevOps and software engineering challenges. Sponsorship inquiries: [email protected]

The Java ecosystem is maturing. The GraalVM high performance runtime provides a virtual machine for running applications in a variety of languages. TornadoVM extends the Graal compiler with a new backend for OpenCL. TornadoVM allows the offloading of JVM applications onto heterogeneous hardware. Juan Fumero works on TornadoVM. He joins the show to talk about the use case for TornadoVM, the design, and the engineering that underlies the system. We also talk about the overall Java ecosystem. Sponsorship inquiries: [email protected]

Investing in enterprise software has become a competitive business. Lots of venture capital firms compete for the good deals at every stage. This level of competition has driven more capital into the early stages. Ed Sim is a partner with Boldstart, an early stage enterprise investment firm. He joins the show to talk about modern enterprise investment strategy and his own varied personal experiences in working at funds. Sponsorship inquiries: [email protected]

Factories require quality assurance work. That QA work can be accomplished by a robot with a camera together with computer vision. This allows for sophisticated inspection techniques that do not require as much manual effort on the part of a human. Arye Barnehama is a founder of Elementary Robotics, a company that makes these kinds of robots. Arye joins the show to talk through the engineering of Elementary Robotics, and his vision for the future of the factory floor. Sponsorship inquiries: [email protected]

The most popular email client is Gmail, the web-based email client from Google. Gmail is dominant, but that dominance has come at a price, namely speed. Gmail caters to the lowest common denominator, serving a large ecosystem of use cases and plugins. This makes for a slow overall performance. Superhuman is an email client built for power users. Rahul Vohra is the founder of Superhuman, and joins the show to talk about the design and engineering of an email client that is made to be fast. Sponsorship inquiries: [email protected]

The Internet Archive collects historical records of the Internet. The Wayback Machine is one tool from the Internet Archive which you may be familiar with. One project you may be unfamiliar with is book scanning. Internet Archive scans high volumes of books in order to digitize them. In today’s episode, Davide Semenzin joins the show to talk through the history of the Internet Archive and the engineering behind book digitization. We talk through OCR, storage, architecture, and scalability. Sponsorship inquiries: [email protected]

Biometric authentication uses signals from a human’s unique biology to verify identity. Forms of biometric authentication include fingerprints, eye patterns, and the way a person walks, otherwise known as gait. UnifyID is a company that builds systems for biometric authentication. John Whaley is the CEO of UnifyID, and he joins the show to talk through techniques for biometrics, and the implementation details that UnifyID has built to turn these into a reality. Sponsorship inquiries: [email protected]

Robotic process automation involves the scripting and automation of highly repeatable tasks. RPA tools such as UIPath paved the way for a newer wave of automation, including the Robot Framework, an open source system for RPA. Antti Karjalainen is the CEO of Robocorp, a company that provides an RPA tool suite for developers. Antti joins the show to talk through the definition of RPA, common RPA tasks, and what he is building with Robocorp. Sponsorship inquiries: [email protected]

After working at VMware for 10 years, Jerry Chen developed an expertise in technology companies. Today, he works at Greylock, where he looks at deals in the infrastructure and developer tooling space. Jerry is an expert in go-to-market strategy and makes investments in technologies that have a good chance at becoming large and profitable businesses. In today’s episode, Jerry and I talk through the dynamics of modern infrastructure investing, including examples of deals such as Chronosphere and Rockset, both of which have been featured in previous episodes of the podcast. Jerry gives his perspective on deal terms, board dynamics, and everything else that goes into a smart investment. Sponsorship inquiries: [email protected]

APIs within a company change all the time. Every service owner has an API to manage, and those APIs have upstream and downstream connections. APIs need to be tested for integration points as well as for their “contract”, the agreement between an API owner and the consumers of that API. Aidan Cuniffe is the founder of Optic, a product built for API change management. He joins the show to explain why there is an opportunity for such a product, and the market dynamics of the space of API testing and change management. Sponsorship inquiries: [email protected]

WebAssembly allows for the execution of languages other than JavaScript in a browser-based environment. But WebAssembly is still not widely used outside of a few particular niches such as Dropbox and Figma. Nicolo Davis works on an application called Boardgame Lab, and he joins the show to explain why WebAssembly can be useful even for a simple application. Nicolo also shares his reflections on TypeScript, Rust, and the future of web development. He talks through the client/server interaction, performance, error handling, and the process of an actual migration. Sponsorship inquiries: [email protected]

Hyperparameters define the strategy for exploring a space in which a machine learning model is being developed. Whereas the parameters of a machine learning model are the actual data coming into a system, the hyperparameters define how those data points are fed into the training process for building a model to be used by an end consumer. A different set of hyperparameters will yield a different model. Thus, it is important to try different hyperparameter configurations to see which models end up performing better for a given application. Hyperparameter tuning is an art and a science. Richard Liaw is an engineer and researcher, and the creator of Tune, a library for scalable hyperparameter tuning. Richard joins the show to talk through hyperparameters and the software that he has built for tuning them.

Anduril is a technology defense company with a focus on drones, computer vision, and other problems related to national security. It is a full-stack company that builds its own hardware and software, which leads to a great many interesting questions about cloud services, engineering workflows, and management. Gokul Subramanian is an engineer at Anduril, and he joins the show to share his knowledge of how Anduril operates and what the company has built.

CrowdFlower was a company started in 2007 by Lukas Biewald, an entrepreneur and computer scientist. CrowdFlower solved some of the data labeling problems that were not being solved by Amazon Mechanical Turk. A decade after starting CrowdFlower, the company was sold for several hundred million dollars. Today, data labeling has only grown in volume and scope. But Lukas has moved on to a different part of the machine learning stack: tooling for hyperparameter search and machine learning monitoring. Lukas Biewald joins the show to talk about the problems he was solving with CrowdFlower, the solutions that he developed as part of that company, and the efforts with his current focus: Weights and Biases, a machine learning tooling company.

As software permeates our lives, there are an increased number of situations where the legal system must be designed to account for that software. Whether the issues are open source licensing, cryptocurrencies, or worker classifications, software overlaps heavily with the law. Just as software is crafted by engineers, the legal structure around software is crafted by lawyers. There are large law firms that have built their business by knowing how to navigate these software and business questions. Mark Radcliffe is a lawyer who has been working with software companies for decades. He joins the show to talk about the intersection of software and the law, which we discuss from multiple points of view.

Code is version controlled through git, the version control system originally built to manage the Linux codebase. For decades, software has been developed using git for version control. More recently, data engineering has become an unavoidable facet of software development. It is reasonable to ask–why are we not version controlling our data? Dmitry Petrov is the founder of Iterative.ai, a company for collaborating and version controlling data sets. Dmitry joins the show to talk about how data version control works, and Iterative.ai, the company he is building around dataset management and collaboration.

Every software company works off of several different development environments–at the very least there is staging, testing, and production. Every push to staging can be spun up as an application to be explored, tinkered with, and tested. These ad hoc spin-ups are known as release apps. A release app is an environment for engineers to play with, and potentially throw away or promote to production. Release apps have been made easier due to technologies such as infrastructure-as-code, continuous integration, and Kubernetes. Tommy McClung is the co-founder of Release App, a company that makes it easy to spin up release environments for your software. Tommy joins the show to discuss release workflows, and his work building Release App.

Chatbots are useful for developing well-defined applications such as first-contact customer support, sales, and troubleshooting. But the potential for chatbots is so much greater. Over the last five years, there have been numerous platforms that have arisen to allow for better, more streamlined chatbot creation. Dialogue software enables the creation of sophisticated chatbots. ParlAI is a dialogue platform built inside of Facebook. It allows for the development of dialogue models within Facebook. These chatbots can “remember” information from session to session, and continually learn from user input. Stephen Roller is an engineer who helped build ParlAI, and he joins the show to discuss the history of chatbot applications and what the Facebook team is trying to accomplish with the development of ParlAI.

Image annotation is necessary for building supervised learning models for computer vision. An image annotation platform streamlines the annotation of these images. Well-known annotation platforms include Scale AI, Amazon Mechanical Turk, and Crowdflower. There are also large consulting-like companies that will annotate images in bulk for you. If you have an application that requires lots of annotation, such as self-driving cars, then you might be compelled to outsource this annotation to such a company. SuperAnnotate is an image annotation platform that can be used by these image annotation outsourcing firms. This episode explores SuperAnnotate, and the growing niche of image annotation. Vahan and Tigran Petrosyan are the founders of SuperAnnotate, and join the show for today’s interview.

Business intelligence tooling allows analysts to see large quantities of data presented to them in a flexible interface including charts, graphs, and other visualizations. BI tools have been around for decades, and as the world moves towards increased open source software, the business intelligence tools are following that trend. Metabase is an open source business intelligence system that has been widely adopted by enterprises. It includes all the common tools that are expected from a business intelligence system: large-scale data ingestion, visualization software, and a flexible user interface. Sameer Al-Sakran is the CEO of Metabase and he joins the show to talk about Metabase’s design, engineering, and usage.

The US Army Cyber School is a training program which trains cyber soldiers and leaders to be adept in cyber military strategy and tactics. In order to teach these skills, the cyber school uses a system they call “courseware as code”, a workflow that allows updates to the curriculum in a reversion-friendly fashion similar to infrastructure-as-code. Ben Allison teaches at the US Army Cyber School and has put work into developing the training program and ongoing lesson plans. Ben joins the show to talk about how the US Army manages curriculum through courseware as code, and the work he has done to improve this workflow over time. Ben is also speaking at GitLab Commit 2020, GitLab’s upcoming conference. You can register for GitLab Commit yourself by going to softwareengineeringdaily.com/gitlabcommit.

Logs are the source of truth. If a company is sufficiently instrumented, the logging data that streams off of the internal infrastructure can be refined to tell a comprehensive story for what is changing across that infrastructure in real time. This includes logins, permissions changes, other events that could signal a potential security compromise. Datadog is a company that was built around log management, metrics storage, and distributed tracing. More recently, they have also built tools for monitoring the security of an organization. Detecting security threats can be achieved by alerting on known security risks, or pieces of information that could be indicative of a vulnerability. Marc Tremsal works at Datadog, and joins the show to talk through security monitoring. Full disclosure: Datadog is a sponsor of Software Engineering Daily.

Dev.to has become one of the most popular places for developers to write about engineering, programming languages, and everyday life. For those who have not seen it, DEV is like a cross between Twitter and Medium, but targeted at developers. The content on DEV ranges from serious to humorous to technically useful. DEV contains a set of features which appeal to a developer community, such as the ability to embed code snippets in a post, but for the most part the entire app is generalizable to other types of communities. Hence, the motivation for “Forem”. Forem is an open source project to make it possible to spin up instances of communities that are like DEV, but for other communities such as mixed martial arts, or doctors. Ben Halpern is the creator of DEV and Forem, and he joins the show to talk about the DEV Community and his long-term goals for what the DEV team is building.

Drug trials can lead to new therapeutics and preventative medications being discovered and placed on the market. Unfortunately, these drug trials typically require animal testing. This means animals are killed or harmed as a result of needing to verify that a drug will not kill humans. Animal testing is unavoidable, but the extent to which testing needs to occur can be reduced by inserting machine learning models which simulate the effects of a drug on the human body. If the simulated effect is negative enough, animal testing doesn’t need to be run, thus no animals need to be harmed. Bryan Vicknair and Jason Walsh work at VeriSIM Life, a company which makes software simulations of animals. These simulations can be used to model drug testing, and change the workflow for drug trials. They join the show to talk through the mechanics of drug testing, and how VeriSIM Life fits into that workflow.

Across a company, there is a wide range of resources that employees need access to. Documents, S3 buckets, git repositories, and many others. As access to resources changes across the organization, a history of the changes to permissions can be useful for compliance and monitoring. Indent is a system for simplifying access management across infrastructure. Indent allows users within an organization to request access to resources, and keeps logs of the changes to who can access those resources. Fouad Matin and Dan Gillespie are the founders of Indent, and they join the show to talk through the application of access control management, and the architecture of Indent itself, which has numerous interesting engineering decisions within it. Indent job opportunities