Danah Boyd, Martin Fowler, Neil Ford discussed on Software Engineering Daily

Automatic TRANSCRIPT

It's a podcast four techies. He's their team of technologists. Take a deep dive into a tech topic each episode. That's pique their interests. It could be. How machine learning is being used in astrophysics succeed continuous delivery. They're always coming across fascinating ways. Technology is advancing and love to share what they learn. Whatever the topic. The discussions are always lively informative and opinionated. The team of co hosts are experienced technologists from across thought works including thought work. Cto dr rebecca parsons and renowned writer and speaker. Neil ford each episode the podcast features guest or to to talk about their particular passion and areas of expertise past guests have included imminent technologists such as martin fowler mark richards and danah boyd. There are a wide variety of episodes on the podcast and a lot of them have to do with software architecture closure digital trust. These are topics that all works has been thinking about for a very long time and there is well-equipped as anybody to talk about them. So whether you want to broaden your knowledge on a specific topic or you just want to immerse yourself in the world of technology theraworx technology. Podcast has something for you to find out more. Just search thought works technology podcast on your podcast platform of choice and make sure to subscribe developments in the machine learning space. The space obviously moving really fast are there any opportunities to leverage all the change going on in the machine learning go system as a product designer for notable like. What do you see as an opportunity. Business wise in in Making better machine learning experiences through your notebook platform or machine learning. There's a ton of great new libraries. They've come up as he gets been a rapidly evolving space but more and more libraries are coming out of the box. Visualization enabled by default and a really natural place to do that. Digitalization is within a notebook. Because you don't have to have some export you're gonna look at or some other system where you're gonna dump the results. Cd needing allies so more and more machine. Owing libraries are actually coming out of the box with support for outputs her jupiter that rendered naturally. And then you get much richer evaluation experimentation pattern for you need much fewer lines of code in order to get a rich representation of the state of the problem. You're looking at or or the state of the result that you're looking at and for the tools that are out there that are that are upcoming or already established. I think that's like a really winning aspect of notebook. Integration that issued looking at solving for that corner of the data user experience that making sure those tools behaviorally naturally within a notebook. Experiences is critical. You might do things like make sure. If they do have associations they they captured cleanly Render nicely you do things like if they don't have visualization but they have david for building visuals. Asian may be automating run in the notebook experience to Additionally triggerfish rendering for the user and making it so that the user can do visualization with less code. I think that's An aspect that's proven Very from lots of people that might be in the surprising fact side of saying that auto viz or even just the the easy viz is super valuable to people. Because you waste so much time trying to fiddle with getting x. axis riot or not quite selecting your data correctly meant having a u y that just rendered automatically the show that i think meets those tools much more useful and are unique in that they have that. Visualize ation built right next to your code execution so you don't have to leave the tool to do that. The production ization of notebooks. I'd like to revisit this. What are the difficult parts of getting a notebook into production and maintaining that notebook in production. Updating it over time for notebooks in general actually reverse it from the other way. I think if you start with a lot of people get stuck on. Hey how would i. Production is a notebook relative to this really sophisticated library with large amounts of devops tooling and de tooling already in place or lots of best practices. The actually it's usually be how those tools in place not look like. Production is anything else you make test. You make sure you have code review and you have a deployment cycle that controls when that note goes to users and beyond that point that the standard execution environment is pretty normal young year in this case usually doing integration immigration tests. And you're usually doing some research. Shaw as promotion to whatever your Production line is for for a exposing asset to user either automated or manual fashion. And when you actually get from the reverse side look at actually how most awards have data usage and you're talking about like things are still kind of not great in data usage tooling alava orange even very big sophisticated. Orcs have a million s three vials. That have some python code number that do their. Etl and there some place in their copy between places. Or they're in some hard drive on prem system and they just have the scripts that are everywhere. And if you come at leveling those like those users have been neglected from tooling the tooling honestly kind of hard to get setup and it's it's one two three context shifts away from the problem that these other users like data analysts data. Engineers are focused on the not focused on best axes around programming tooling. The focus on getting data results of their to their business to analyze. So they're looking to care much more about the like the mentions of the data that working wiz and like the consistency of it and tertiary concern to things like code quality your virgin control and they're gradually growing into using those tools in their things to make them smooth. I don't think actively about using those aspects but if you start from that. Pov of a user had a script someplace that stash. I don't know where and you on that for your production and then that person moves teams or the organization than your etiam pipeline stops working. One day you have very little visibility so your production is asian. Script was pretty poor. If you think of how it notebook could be used in the same way where i developed my notebook personal space. I a schedule with it. And then i kind of walked away or running every morning on monday. Manley go run my no. This is a more common pattern and people care to admit so. I think when you enter our production is asian. You really wanna make it easy to get away from those patterns with user having sink a lot about what they're doing so one thing yell making visible versions of automating the scripts. They're doing to be to build some sort of linear actual urgent control system to get visibility into changes over time without the user having to change their workflow substantially also making the tooling for how. You're interfacing one thing.

Coming up next