Code[ish]

Julián Duque is a Lead Developer Advocate at Salesforce and Heroku, and he's continuing a previous discussion with some members of Respeecher. Respeecher has created AI software which works within the speech-to-speech domain: it takes one voice and makes it sound exactly like another. Dmytro Bielievtsov, its CTO and co-founder, explains the practical uses of the software, such as re=recording the lines of an actor who is unavailable, or bringing historical figures to life in a museum.

In terms of sophistication, there are quite a few speech ML models already available on the Internet. The best source of audio to duplicate the speech patterns of a famous person is to grab an audiobook and pass it through one of these pre-existing models. But these models produce outputs which are poor in quality. That's one reason that speech-to-speech is hard to fake. The variation in our mouths and speech patterns, not to mention the emotive qualities, make the process of creating duplicate voices extremely difficult to pull off. One way Respeecher data can't be faked is by the fact that they position themselves as a B2B business, dealing with studios and other large estates which have access to immense amounts of hard-to-acquire sound data. The likelihood of another entity abusing a well-known voice is close to none. Another feature is that the audio is watermarked. Certain "artifacts" are embedded into the audio, which are imperceptible to humans, but easily identifiable by a computer program.

There's a consortium of several companies working on synthesized media who strategize in Slack on various ways to keep the tech from being misused. As well, Dmytro believes that there needs to more investment in education, to let people know that such technology exists, and to therefore be a bit suspicious with media they encounter online.

• Respeecher is a speech-to-speech voice cloning software
• Google's DeepMind and Tacotron are two popular AI/ML projects
• Module empowers the future of identity, online safety, and communication through voice
• Previous episode: 98. The Ethical Side of Deep Fakes

Julián Duque is a Lead Developer Advocate at Salesforce and Heroku. He's joined by Alex Serdiuk, the CEO of Respeecher. Respeecher has created AI software which works within the speech-to-speech domain: it takes one voice and makes it sound exactly like another. Alex rejects the premise that all deep fakes--that is, pictures and videos generated by AI--are inherently evil. He considers tools like CGI and Photoshop to fall within the realm of synthesized media, which helps artists create content. He positions Respeecher within that same mileu.

Respeecher has been working with Hollywood studios for some time. It removes pressure from actors who are unable to rerecord lines. It's also been used in situations where actors need to sound much younger, a visual-audio process called de-aging. In the future, applications of speech-to-speech work could also be used in museums, to provide a new dimension of history for audiences.

Of course, Alex recognizes that the main issue with deep fakes is not their existence, but their inability to be detected. To solve this problem, Respeecher watermarks its audio, to generate inaudible metadata which can nonetheless be analyzed to show whether a particular recording was faked. He also believes that more people need to be educated that synthesized media exists. Something one sees or hears might not be real, because technology is getting more and more advanced. We should all be mindful about the content we consume.

• Respeecher is a speech-to-speech voice cloning software
• Samsung Next provides an overview of the synthetic media landscape

Jacob Silzer, Trusted Security Director at Heroku, is co-hosting this episode with Trey Ford, VP of Platform, Trust, and Strategy at Salesforce. They're sitting down with Tim Panagos, CTO of Microshare. Microshare began as a platform to aggregate wireless sensor data from IoT devices. Much of their use cases were for smart offices; for example, seeing how occupied a conference room was, whether a hot desk was open, and when a particular area had been cleaned. For hospitals, their platform monitored physical equipment, such as patient beds and medicine carts. These devices ran on a LoRaWAN network, which doesn't communicate through Wi-Fi, making it ideal for secure locations.

Then, COVID hit. Suddenly, information from these devices, where were useful from a logistic and compliance standpoint, became extraordinarily important. By tracking the amount of people in a room, the amount of times an area was cleaned, or the movements of an unknown carrier, whole industries can monitor their overall health efficacy, comparing their performance not only with their past, but also, with other similar businesses. Tim gives the example of an airport here: one could potentially see how their cleaning efficiency is compared to other airports around the world. Part of the success of their platform is their reliance on the blockchain. Businesses can monitor not only how their data has changed over time, but they can look back on any block and identify what factors may have contributed to a decreased in performance.

Tim notes that the platform is not bulletproof. For starters, tracking equipment requires people to remember to add trackers; a simple error, to be sure, but one which is frequently forgotten as new crash carts are added to quickly meet increased demands. But more urgently, no one really knows how the future will change in response to the knowledge we only currently have. Microshare was only able to pivot because they were wise enough to recognize that their business and their software needed to be flexible. They essentially threw out their 2020 roadmap in order to focus on tracking clean air and open spaces. The best way to move forward is not just to collect more data, but to also add context to it, which will create meaningful value.

• Microshare is a data leverage platform utilizing IoT devices
• LoRA Alliance is a non-for-profit group consisting of companies committed to LPWAN

Greg Nokes, a Master Technical Architect with Heroku, interviews two members of Yobota, a banking systems provider: Ammar Akhtar, its CEO and co-founder, and James Maidment, the head of Technical Operations. The financial industry is heavily regulated. As it stands, it was only until about 2016 that the UK (where Yobota is based) gave favorable guidance for vendors to operate in the cloud. As a service provider, the banks that use Yobota are audited by the Financial Conduct Authority. As part of that audit, every single deployment performed over a year is examined. Regulators select a random set of them, and Yobota has to demonstrate that they know who was involved in the release, and precisely which services were affected. Thus, their entire shipping process is revolved around meeting this regulation goals. They're an integral part of the company, just as data security and uptime availability are.

The platform is designed in such a way to both evolve quickly and quickly perform safe deployments that are observable. Unlike other startups, Yobota has decided to invest in a sysadmin team, in order to split the organization between people who develop features and people who manage their compliance. For example, as the company grows, they've found that active hands-on management of permissions has been a valuable investment. Different groups need access to staging environments versus production environments; and, with over 300 apps on multiple dynos, access to resources needs to be carefully configured.

This is seemingly slow shipping process is advantageous for two reasons. First, meeting compliance is the law, and flirting around that has tremendous consequences. But second, and more importantly, Yobota also provides fake environments for their engineers to develop around. They're able to give developers the ability to experiment with their platform in a safe way; should they choose to advance a feature into a production environment, a different team is able to address what needs to be done to meet the needs of that regulated environment. James suggests to other companies working in these sorts of industries to consider compliance integral to the way their systems operates, and to think about concerns upfront, in advance of working on any feature.

• Yobota is a core banking platform that allows financial institutions to launch innovative products in a fast and reliable way

Corey Martin values storytelling. It's just one way developers can share their experiences in order for others to take lessons. To that end, this episode takes a close look at production issues from two different applications to examine what went wrong and how it was fixed.

Meg Viar is a Senior Software Developer at Nomadic Learning, an e-learning platform. One day, they noticed that, for a certain group of users, a column of information in their database row was nulled. It didn't look like any user--either internally or externally--intentionally changed these values, and there hadn't been any new code deployed in days. The only clue was that the data was all changed at the same time. It turned out that a weekly cron job was deleting some data on an in-memory list. However, the database ORM they use also overloads the delete keyword, and was actually deleting the production data. Restoring the data from a backup was easy, and reworking the code to not use the data was a quick fix. However, going forward, Meg and her team came up with several ways to adjust the process around code changes like this from occurring again.

Brendan Hennessy is the co-founder and CTO at Launchpad Lab, a studio that builds custom web and mobile applications. One of their clients is an SAT/ACT test prep app, and students complained that the app was extraordinarily slow. Brendan was accustomed to seeing such feedback on testing days, when heavy volume brought added strain to servers, and they accounted for this by increasing capacity. But this was different: there weren't any tests scheduled during the period. Instead, one of their own services was inadvertently DDOSing an endpoint, expecting a response; when one didn't arrive, it just kept making requests. They reworked this code to make a request once and simply wait for a response without trying again. In the future, they committed themselves to doing more in-person blitzes of new features, since issues like this only arise after multiple users use the app--something automated tests have trouble simulating.

• Nomadic Learning builds digital academies
• Launchpad Lab builds custom web and mobile applications for startups and established businesses

Becky Jaimes is a product manager at Salesforce. She's interviewing Wesley Magness, the founder of ElectricSMS, and Melanie Plaza, the Head of Technology at AE Studio. ElectricSMS is a service to help consumers manage their various subscriptions, whether that's recurring orders of dog food or monthly boxes of snacks.

ElectricSMS started as a project within AE Studio's incubator program, fitting in with their ethos to empower people through technology. AE Studio is a bootstrapped company that works with clients by offering development, data science, and design help to enable startups and enterprises to build technology products that helps humans, not profits. Their philosophy is to build products that treat their users well, so that the business can grow much more healthily than if you trick users into purchasing something they didn't mean to, or otherwise making it difficult for them to cancel. ElectricSMS is an example of how those principles came together in a collaborative partnership.

For listeners interested in starting their own business, Wesley suggests finding a customer before working on something. You don't need an actual user to work with, but rather, a customer or industry who you know would benefit from what you built. With services like Heroku and Stripe it can be very easy to build an MVP from the ground up. Melanie concurs, and suggests talking to people who might want what you're making, and what their problems and needs are. Research what other solutions have and haven't worked. Startups can also reach out to AE Studio for direct consultation on taking their product to the next level, and potentially join their incubator program as well.

• ElectricSMS helps users manage their subscriptions through their phone
• Effective altruism movement advocates using evidence and reasoning to determine the most effective ways to benefit others

Corey Martin, a customer solutions architect at Heroku, interviews Ariel Assaraf, the CEO of Coralogix, a platform that helps companies get a grasp on their log data. All too often, logs are considered as only a useful debugging tool. After receiving an alert around high resource usage or an elevated error rate, a developer might check their logs to see what caused the issue. But Ariel argues that this is too late to investigate a problem; by visualizing and alerting log data, you can figure out production problems before users encounter them.

Metrics, in other words, are a lagging indicator, while logs are a real-time representation of how your code is really performing. One way to reconcile these two is to aggregate log data and funnel it into other long-term metric storage. This would allow you to see longer term trends. Ariel provides a scenario where log records appear in groups, such as a user purchasing a product, followed by an API call to Stripe, and concluding with an email notifying the user. A platform like Coralogix can automatically identify that the three logs arrive together within a certain time frame. If, for any reason, one of these steps fails to log, then a notification can be set up to notify the team to proactively investigate, rather than a customer writing in to report an error.

For an organization to beginning using logs as time-series data, Ariel recommends three things. First, a unified log format, which could be something structured like JSON. These can be generated by a middleware service. Next, a shared understanding across teams on the severity with which to log a message. The final step is to set up an alerting policy; not only which types of alerts to create, but also where they go, such as Slack, email, or text message. After that, you can begin to incorporate your logs into your monitoring processes.

• Coralogix is an observability platform for logs, metrics, and security

Anand Gurumurthi, a Director of Engineering at Salesforce, is joined by Marcus Blankenship, a senior manager at Heroku Salesforce on the Runtime Networking Team. Their topic for this episode is to provide career advice for experienced engineers looking to advance their career.

There are several all too common scenarios individual contributors face, which Anand and Marcus discuss and offer their perspectives for. These include learning how to ask overcoming bias on your own work and figuring out how to better assess your strengths and weaknesses. In addition, Marcus and Anand discuss getting over the fear of asking for help. They agree that showing an active interest in finding out how a system works is a good way to show that you're engaged in your work and want to grow.

In summary, confidence in the workplace is the key to getting what you want. Whether you're looking for a promotion or want to better understand a problem, you are in charge of starting those conversations. In situations where you're looking for guidance, ask for specific feedback from people who want to help you.

Chris Castle is a developer advocate at Heroku and Salesforce. He is joined by Carter Rabasa, the lead organizer of CascadiaJS, as well as Julián Duque, a developer advocate here at Salesforce/Heroku who organizes NodeConf and JSConf in Colombia. Carter shares his first experiences at a tech conference, finding it to be surprisingly intimate and a great community of well-intentioned web developers that wanted to learn. He was inspired to start CascadiaJS, a JavaScript conference situated in the Pacific Northwest. Over time, he realized that it's the people and the networking opportunities that really makes CascadiaJS special.

When COVID-19 made it clear that in-person events would not happen for 2020, he and his team struggled to figure out how to put on an event that their community would love. It required them to imagine a future where software to support their vision didn't exist yet. They became certain that the event would need to learn how to be virtual for a long time. They accepted this challenge, and set to work building a conference model that they felt was interactive and immersive. There was just a tremendous excitement and enthusiasm to see if they could do something that hadn't been done yet.

Of course, they stumbled in several ways; there were issues sending swag to customers, for example. Still, there are many reasons to keep the virtual conference format. For one, it's more accomodating for people with physical accessibility issues as well as attendees all over the world. There's more flexibility in the timing of events, where speakers can just play their sessions one after another; attendees can hop between different workshops and talks at the click of a mouse. Julián agrees that CascadiaJS' hybrid format of a recorded talk followed by a live Q&A was great for engagement, as speakers were chatting with viewers as their session played. Overall, Carter is excited at future conferences having a serious virtual component to them.

• CascadiaJS 2020 went virtual-only for the first time
• We Need Better Virtual Dev Conferences article
• NodeConf and JSConf in Colombia are just two events grappling with the pandemic

Dr. Mireille Reece is a practicing clinical psychologist and, along with Adam Stacoviak, editor-in-chief of the Changelog, they run Brain Science, a podcast exploring behavior change and mental health. Chris Castle, a Developer Advocate at Heroku, continues a previous conversation with them on how to take care of one's mental wellbeing. Their discussion centers around the neuroplasticity of the brain, how new habits can be created and formed.

One of the best antidepressants is exercise. Adam advises that it's not necessarily about physical fitness, but rather that your brain really needs that motion. It can serve to provide one with a different perspective. As well, activating any of your senses--smell, touch, and so on--can bring on a calming effect. Other suggestions include meditating, which can bring on a greater sense of self-awareness, and journaling, which can be helpful in tracking your thoughts over time.

Dr. Reece advises that keeping tracking of your moods can help hone in on the source of negative emotions. It's all about cause and effect. For example, if you eat a lot of fried foods, it might taste good, but ultimately, there are long term health affects associated with it. It's not that our actions are bad or good, but about recognizing the effects of them. Once we hone in on unfavorable consequences, we can start the process of adjusting our behaviors away from their sources.

• Brain Science is a podcast exploring the human brain to understand behavior change, habit formation & mental health
• Headspace and Calm are two apps which can help with mindfulness
• Day One is an app for journaling
• Essentialism is a book on "pursuing less" which Adam recommends