A Collaborative AI Development Tool with Tommy Dang of Mage

Machine learning and AI has grown from novelty technology to essential requirement. But unfortunately, most smaller tech companies don’t have the expertise among their staff to implement these solutions—and sometimes don’t even understand how they work.

This is why Tommy Dang co-founded Mage, which offers a collaborative AI development tool to help companies stay on the cutting edge.

More information:

Tommy Dang grew up in San Jose, CA and studied interdisciplinary studies at U.C. Berkeley. After graduating, he taught himself how to code and started building web and iOS apps. In 2013, he co-founded a short term rental marketplace startup called OnMyBlock that helped students find off campus housing.

After 2 years, they shut down the company and Tommy joined Airbnb. Shortly after, he was recruited to join a small team led by Brian Chesky. On this team, he helped build Airbnb Experiences and launched it to hundreds of millions of people around the world.

In 2018, he created an internal low-code tool at Airbnb, called Omni and built the team from the ground up. This tool was a combination of Squarespace, Mailchimp, and Google Ads rolled up into a single platform.

Hundreds of developers across Airbnb have built new features on top of the platform and over a thousand Airbnb employees have used Omni to launch landing pages viewed by millions of people, send billions of emails, and create thousands of promotions.

After over 5 years at Airbnb, Tommy left and started Mage to help equip product developers with accessible AI technology.

Mage is a collaborative AI tool for product developers. We’re making AI accessible to front-end developers, backend developers, and native developers at SMBs. Mage will make AI technology so accessible, a mom-and-pop shop can harness its power.
We believe in a world where every business, small and large, can create transformational products for their customers.

Our mission is to equip developers with accessible AI technology so they can deliver magical experiences to their users.
Mage is the Stripe for AI. Stripe made it easy for developers to integrate payments into their apps. Mage makes it easy for product developers to integrate AI into their apps.

DISCLAIMER: Below is an AI generated transcript. There could be a few typos but it should be at least 90% accurate. Watch video or listen to the podcast for the full experience!

Tommy Dang 0:00
One of the models we have there work, it’s better to be in the wrong place with the right people than in the right place with the wrong people.

Alexander Ferguson 0:14
Welcome to UpTech Report. This is our Applied tech series. UpTech Report is sponsored by TeraLeap. Learn how to leverage the power of video at Today, I’m excited to be joined by my guest, Tommy Dang, who’s based in San Francisco. He’s the CEO and co founder at Mage welcome Tommy, good to have you on.

Tommy Dang 0:32
thank you so much, super excited to be here. I love what you’re doing.

Alexander Ferguson 0:35
So tell me from what I understand what you’re doing at mage is, you’re a developer by trade like you, you know what they think what they understand, coming, though, spending a lot of time machine learning. Developers are used to taking concept to completion, like be able to handle the whole thing. And what you’re trying to build is developer tools to help them use machine learning Bing, if I understood you had shared this analogy, Stripe for machine learning, development and you feel machine learning is becoming table stakes in a lot of the cases, and you’re not focused on no code solutions. It’s really still to the developers out there, helping them create a this those maybe their product, or small or midsize business. So if you’re in there, and you’re listening or developer out there and a small and midsize, but you don’t have a machine learning expertise. This is a platform did I get that correct?

Tommy Dang 1:26
Yeah. That was spot on. Definitely the stripe for for machine learning and helping developers be able to integrate machine learning even if they haven’t worked with in the past.

Alexander Ferguson 1:36
Okay, so let’s let’s go back and let’s let’s hear the journey because you you started this? What When did you jump into this? With mage? How long ago was this?

Tommy Dang 1:44
We founded it in December 2020.

Alexander Ferguson 1:47
December of 2020. So you lots of energy excitement is like, there’s so much to uncover and expands. But before this, you you were an Airbnb for for six years,

Tommy Dang 1:59
Zachary? A little over five years, little over

Alexander Ferguson 2:03
five years. Okay, almost six. And you were doing machine learning development there.

Tommy Dang 2:09
We always doing a lot of software engineering growth, the whole journey, there’s a whole journey. Maybe I can share with you.

Alexander Ferguson 2:16
That for me take take me back. What was that? Like? Yes, yes.

Tommy Dang 2:19
So I joined early on in 2015. And shortly after I joined our CEO, Brian was putting together this small close knit team, what magical trips. And this team would go on to build, launch and grow what we all know today as the Airbnb experiences business. And I was fortunate to work very closely with him. So all building a product, a business from the ground up in scaling to millions of users. Now, while I was on that team, we were responsible for growing this team, meaning we would bring top of funnel users reach out to users who’ve never used it before, get them to book it, leave reviews, and come back. So we were the growth aspect of it. And while on this team, as developers, we had to scale our efforts. So we were building lots of landing pages, sending emails, putting doing text messages, push notifications. And we found that we were doing this over and over and over. And so we started building scripts and tools for ourselves. But over time, we saw that other developers, other companies were doing the same thing. And other non developers were trying to ask developers to help them do the same thing. So my now co founder and I at Airbnb in 2018, we started this internal SAS developer tool, it’s on me think of it as a Google ads, but Squarespace plus MailChimp, all rolled up into one. And so we started, we started this tool from the ground up, built it out, it’s a low code tool for developers and non developers to build any pages without having to write code, ship tons of email. And through this process, we got to work with lot hundreds of developers across all the different teams. And what we found was that a lot of the developers that were working with, you know, they, they have great ideas on how to use machine learning, they knew where in the product, it was being used, where on the website, but when it came down to actually working on these projects, and funding these projects, they always needed relied on a data scientist on another team, or it was gonna take too long to build and, and I saw this hurting innovation, excitement, and potentially lots of revenue. And so we started looking around at the available tools that there must be some tools today. And you know, there are there are two Airbnb actually built some internal tools. There’s some really great internal tools that Airbnb built on the machine learning infrastructure team. But you know, what, on one hand, how many developers knew how to use this tool, because tools today are built for data scientists, AI researchers, machine learning engineers, by other data scientists, so the experts already but people who are doing on a daily basis already With the available tools, there’s no available list today that cater to developers need or developer needs. Engineers back end, native. And so we set out to build mage, the stripe for AI stripe made it really easy for any developer to process payments and integrate it into their apps mage makes it really easy for any product developer to integrate models into their ops.

Alexander Ferguson 5:23
This this concept of making it more accessible, easy for those that are machine learning experts or data scientists. Do you? So your assumption is that not every developer will need or should need to learn the ins and outs of machine learning and data science?

Tommy Dang 5:42
Yeah, absolutely. This is where from our insights is that developers in the machine learning aspect, applying machine learning, the biggest parts that are involved is knowing where to get the data, or actually first starting to begin to get asking the right questions. It’s how do I take more users onto our platform? How do I get users to buy more? So that’s asking the questions. And developers are really great at asking these questions, because they’ve worked so closely to the product and the ability, and my developers actually build products for you. There’s, and then the next part is asking yourself with this witness question, what data do I have available that can help answer this question. And again, developers are perfectly poised for this, because if you’re building out a product, you have to figure out the database schema. So you already understand the data that we’re collecting from the users the forms. And if you want to actually collect user activity data, who’s implementing a ticket, think of this aptitude, who’s building in the amplitudes, and the mix panel who’s integrating that and collecting those data developers are, and everything in the middle, preparing the data, building up the model, managing pipelines, all that is, I would say, doesn’t involve that much creativity, it could actually drain your creativity. But at the end of it, after you’re all done, what’s important is applying the model, you might just have this model that arbitrary binary file sitting around, but actually flew integrates it into the user flow. At the end of the day, the developer still does. Even whoever builds the model, the developers still need to go into their app. And now all them all at all, do some if else statement to say no, if logged in, do this else use the model. So developers are still involved so much in the office already. And the one the one or two areas that they’re not involved in. It’s because it relies on the data scientists who, who knows good can set up that pipeline can set up the Jupyter Notebooks, knows all the machine learning algorithms and all that. So that’s where maze comes into helping the developer with that aspect of it.

Alexander Ferguson 7:53
You mentioned that machine learning is table stakes. And there are more and more solutions and tools coming out. But what you’re saying is, you saw a gap, where it was targeted people who aren’t machine learning experts, it’s like those who are machine learning experts, there’s more and more tools available for them. But the average developer now for those who are listening, if if you’re not a developer, you may remember Larry, this is this a lot. But I think it’s powerful for developers, but actually taking a step back more than just the business model here for a moment, because I like to look at both in a lot of our conversations of the technology. How does it work? Why is it different? How’s it solving a problem? But also the business? Like how are you making this work? What’s What’s your business model? You just planning? Is it? Is it a usage fee? Because you’re going for more of a product lead growth on this? Correct?

Tommy Dang 8:44
Yeah, absolutely. And so we also touch up upon the table stakes, it’s because there’s a misconception that only large companies need machine learning. Only large companies have enough data to train models. Now there’s a right assumption that only large companies have their available skill set. And that’s mostly true. But the other two, but unfortunately, that’s not the case. There are steep days series, AC companies that have a trove of data in 2020. And beyond. Data just comes so abundantly. And when you’re building out a company, you actually build out your developer team first. And then only later on, do you bring on data scientists? There’s actually multiple types of data scientists, people, companies don’t even bring on GSI just to just work on machine learning. First, we build on it. We have data analysts, and then other types of data science. And so with me, we can equip me with that with their existing developer team. Are you talking about business model? And that’s why we are building for product developers at small to medium sized companies. Because not only are the existing tools today, built for data scientists by these methods, they’re also tailored for large enterprises. They’re solving needs for organizations with hundreds of developers or are hundreds of data scientists and they have a different set of needs and priorities. And so that’s why we’re making AI more accessible to these other small companies. Now, you talked about pricing model, we talked, we initially were working off of usage based pricing, where how many API calls you make, how many predictions you made. But what we found was that one of the biggest values we add to our customers, is the ability to try things out quickly see what works, and not have to waste a ton of resources, and ramping up and build spend so much effort in a model and not knowing if it works at the end of the day. So we wanted to shift more to a pricing model that incentivize our customers to just build, build, build, try, try try and not be afraid to fail, because it did, you don’t have to put in so many resources into a certain model. And that’s where I believe a lot of companies may or may not be extracting enough value out of out of machine learning. Because they don’t have the capability to just live filmy things fairly quickly. It’s like throwing spaghetti at the wall, it’s still there to help you know how they say fail fast, Lean Startup, build a bunch of features, see what sticks. Well, that’s just not the case in machine learning. Now, you know what, it takes a day with the tools mosaic, you got to say, Okay, let’s just pick lunchtime, this Steven work? And then maybe iterating it? How can we take the punt development mindset that the hacker mindset and apply to machine learning?

Alexander Ferguson 11:35
You had get given credence earlier that developers are lazy? No, they don’t want to do things over and over again. They want to try iterate try this. And, and really do you see is like the ability here for machine learning is like, I don’t want to have to learn all that spent a long time figuring it out. I just want to test it. Play around with it. Is that right?

Tommy Dang 11:55
Yes. Yes, absolutely. You know, if you look at development, we have great tools for DevOps for hosting front end for for testing. naamyaa, we’re builders, product developers or builders, you know, it’s great, we sometimes peek under the hood. But as long as it works, we’re good. We just want to create value for the end user. And so that’s how but cool like mage really helps helping developers create more value for the end user.

Alexander Ferguson 12:22
Tell me I can send you have a lot of passion. Yeah, a lot of energy for this. And I feel it. I love it. You were this is the first company that you’ve led, you found it. What learnings are you taking from your time at Airbnb to now apply into this new endeavor?

Tommy Dang 12:43
No, actually, I before Airbnb, I started a company with a friend in San Diego. I might have not I might have not mentioned it before. But we started this, this two sided marketplace company and on my blog, we help students find off campus housing, we raised a seed round a Series A, we ran it for about two years grew the team about 25 people eventually shut it down returned to almost all of what we raised. And that’s why I chose to join Airbnb. We were working in the two sided marketplace and it’s sort of real estate and short term rentals. And then Airbnb at the time. We really love Airbnb for its design, engineering culture. And if we think about marketplaces, who was the best one at that time, Airbnb, we then I joined Airbnb was really awesome. I’ve learned so much from that start, and I applied a lot of our hustle, the you don’t really agree every one of their core values is the a serial entrepreneur, and serial spelled like the food serial, but it’s also meant to be a pun for serial doing it multiple times. And so you know that that value is saying you do whatever it takes go above your goal, you know, don’t be a role define responsibilities to deliver so much value and make a wow experience to the end user. And so I learned that firsthand, working on my first daughter. And so I brought that into everything. And then I was able to really exemplify that core, that core value that helped me also be able to contribute so much in building the Airbnb experiences business. And that it fits that I have at my first job also helped me and my now co founder create a startup inside of Airbnb, the AMI tool I told you I share with you about in my experience, they’re building out a team of 25 team members, cross functional, almost 20 engineers on that team, you know, building out the tool from the ground up. I did marketing and sales that you think like what is marketing and sales internally. You know, it’s tough. You got to fight for team’s attention. You got to you got to get in there meetings. Yeah, I have walked around the pitch deck. I pitch people so many times. We have created our own marketing material, your own customer support channel. We use Slack at MTV, but we don’t because it’s more channel we tend to 24 because we had we, everybody had employees all over the world, right? And so all of the world, people were using article on me. So we’ve had a 24/7 customer support. So we, you know, we’re building this internal startup inside of Airbnb for over two years. And so took a lot of that team building the culture building, Div product development process, you know, iterating, all that experience. And we took it also to mage and we took all of our experience building tools for developers, you’re working so closely with developers and knowing what what drives them? What parts of the tools do they love using what parts of the like coding or parts of the don’t like coding? And so we took all those learnings and growing it into major just doing doing bigger?

Alexander Ferguson 15:45
I really appreciate the clarification. I’m sorry, I missed that. That previous startup you had I mean, and now also, the internal startups really had two stars before this one. Let’s go back to the first one. On my block, right. Yeah. And that was often it’s failing that we learn even more than winning. Sometimes we don’t like to fail, who likes failure? Everybody says that is wrong. But help me understand learn from you. Like, what is the takeaway that you took from that one? That was that you’ve applied now? Going forward?

Tommy Dang 16:20
Yes, yes. So some learnings we learned from that was the importance of a team, we checked with one of the modules we have there with, it’s better to be in the wrong place with the right people than in the right place with the wrong people. And we just really built a great team there. And we, we, we knew that the power of building culture, establishing conditions, establishing your quirky, quirky things, and carrying those through and working very closely, and then hiring, directly hiring right, not not, not trying to grow too fast before establishing your great team working relationships bolter, as well. We also learn how to, we also learned how to really understand users and really build for the user and really have empathy and compassion for the users. And one thing that we learned from there, and we didn’t have a saying for it yet, but actually learned more of it at MVP, but really use your product day to day out. And it was really awesome. Because we we had students on our team, and we had interns and things like that, and before our co founder was just out of college, and so we have the opportunity to really use our product. And that makes a huge, huge difference. And carry that over to me that Brian would always say walk the park. And because Disney is Walt Disney is a huge influence. And he was a visionary before his time, and she would always walk Disneyland park before it opened in there. And so we always now say, lock the park. And that can mean use your service user tool a day in day out, just really put yourself in the shoes of you’re in the shoes of your users if even if even if you can’t figure out a way. And so we did that really well. And on my blog we had we would always go to campuses, we would have this campus ambassador program where we would bring on students to help us grow the demand the supply, we’d bring them into the office, we do events together. And so we really ingrained that into our our culture and even though large amount of the team work college students, but they were our end users, we would do everything in our power to reach neck and stay close and in tune with our end users.

Alexander Ferguson 18:52
They learning for always falls forward of being able to get from the first venture and then yes, definitely within side Airbnb. And now with your new venture. This is still fairly recent. You did get some VC backing. Is that correct? On this one?

Tommy Dang 19:09
Yeah, no, we did raise a pretty good seed round in January. But we haven’t announced it yet. And we’re actually doing a press release next month talking about the raise. So yes, we did raise. We are venture backed by some notable Great, awesome investors and some really great angel investors will be announcing that in a press release next month.

Alexander Ferguson 19:36
By the time this airs, it’ll probably have already not and if not, we’ll make sure it does air in time that this happens. But it’s it’s knowing that other people see this the same trajectory and the opportunity that exists. What are you most excited about? Because it’s still it’s still very soon. There’s still lots to work on. But what is on your roadmap and what’s coming up and Are you excited about?

Tommy Dang 20:01
Yes. So what’s on our roadmap, it’s now we have paying customers, we have developers using your product. And so what’s on our roadmap is making it more self serve, making it easy, helping developers use it even better and faster. And the idea is that we want developers to be able to come and not be overwhelmed. Because there’s a lot of tools out there, there’s just too much. And it sounds funny, as a developer, who is technical, who can code but you’re just like anybody else, he will want simplicity people want guys you want to be No, I can’t do a thing, just help me get to that thing really quickly, really fast, without having to pull my hair out. And so it’s just really refining the product. And your design is actually our competitive advantage. We value design just as much as technology. And design for us isn’t how things look, it’s how things feel in work. And that’s ingrained into our DNA. And the technologies. That’s not enough. The days, you know, everybody has good technology, and especially in this space, everybody has great technology. But the Martin, the leader in this space, we the one who has not only the most powerful tool for the most intuitive and simple tool. And so we are so focused on refining that refining that user experience, we would design and build something really awesome. And then when he hits users, and he, wow, they clicked on none of those. And so we would just, we would just scrap it and redo the whole thing. And that’s okay for us. Because we, we work on tools that we use tools in the past where, where people would build something, and they just gave us to say, Okay, if you don’t figure it out, you know, here’s more guides throw guide at them. But when’s the last time you use an Apple product that came with the instruction manual, but millions, hundreds of millions of people are experts at using it. And so we really believe in, in this aspect. Now, another aspect, another differentiator that we believe in, is, is collaboration, we actually call Co Op similar to when you play games, you play a game, where you team up with other teammates to accomplish the adventure quest. And so collaboration is ingrained in everything we do. You know, there’s that saying, it takes a village to raise a child? Well, in our eyes, it takes a cross functional team to raise a model. And we’ve seen it firsthand. Traditionally, you’d have a data scientist would go out, find all the data, come up with the use case, build the model. And then hey, after six months, they’re handed off what we’ve seen what works, the best is this, it’s, you know, talking the product, talking to design talking to marketing, talking to customer service, talking to the developer and talking to analysts, and getting all that input into you know, can you have this data, or you have this data means nothing without this data? Or you don’t, let’s actually think about using this model differently, or you have, we should optimize for getting the most answers, right, versus, you know, catching all the bad actors, for example. And not one person has all this information on one person can drive it, but Right, willing to

entice case, an invite others in to what contribute in a meaningful way in a timely way, will revolutionize the way and the quality of the models that gets shipped out. And so collaboration is really big. And so we’re constantly building out new ways to collaborate within within me. And now, what’s also really cool thing that is coming up on our roadmap, is this concept that we call open world. Similarly, we talked about Co Op, where you team up together on a team and you go and, you know, level up together, will open the world is it concept of where you three, you’ve pre formed, you go out and you explore on your own, and you might meet new people out in that world. And that’s meeting new maze developers in on the platform in a community type aspect. And they’re Imagine being able to get get help from others, or contributing and sharing some types of transformations. You’ve done in your data, some new techniques that you’ve prepared your data, or maybe you build a product recommendation model, you can show Hey, these are the ways I’ve improved the model. And if you’ve seen our product, we actually in a co op mode, you can see how someone has changed their data, you can see the history of it. And also with the mono you can see the history of the things that anybody has improved on it and the incremental improvement. So those are things that you can acknowledge, share really easy when you don’t have to share sensitive data, I can, I can have my sensor data and say that, you know, all the numerical columns. I, I aggregated the sum over seven days, five days in 18 days, and for that Say, e Commerce Data Now what really well, you can go into the open world and say, Oh, you’re on e Commerce Data, check out this transformation I did, you should try that on your data, it helped us a lot. So this open world, because, again, our goal is to help developers advancing their careers, expand their skill sets, and mastered their craft. And part of that is, you know, becoming proficient machine learning. So we in the open world note it’s to, it’s to better serve the community and our users,

Alexander Ferguson 25:31
you definitely have a greater vision than just, here’s another tool for solving this person’s problem. Or here, it’s trying to move and make a movement within the developer community, to making machine learning possible. This is exciting. I could definitely, as I said before, see the passion in this for those that want to learn more? Ma It looks like it’s going to click button to get access. So when it when are you planning on opening it up even further?

Tommy Dang 26:01
Yes, great question. We are going to launch out a private beta. So general availability, January 11 2022. And so right now, what we’re doing is working with early paying customers, and onboarding developers who are who are requesting access, but you know, just come to me age, get access and enjoy our community channel, and we’d be happy to onboard you and show you how to use it. Because, you know, we want to, we’re taking a break. We’re doing things that don’t scale, we’re taking an iterative approach to building out the core community, the core, because we, we want it to be a group of initial group of thought, really passionate people who who are hungry, who just want to learn and grow and who want to teach others, you know, one of our core virtues, is give people power ups. And that means we, we give honest, immediate feedback with positive intent. We do everything in our power to encourage others. And we see the potential of others, we don’t see people for who they were or what they’re doing with who they can become. And so we’re building the same way. We’re building out our team, we’re building our community, the same way of people who want to help others grow and become better.

Alexander Ferguson 27:18
I mean, thank you for sharing this passion of yours, the product that you’re building, and the vision that you see, this has been a great conversation.

Tommy Dang 27:29
Thank you so much, Alexander. Thank you so much for having us on here and loving what you’re doing and stuff.

Alexander Ferguson 27:38
Absolutely. For those again, want to check more, go to Can sign up of the private beta or when it launches open more, you could be able to get access there. And we’ll see you all on the next episode of UpTech Report. Have you seen a company using AI machine learning or other technology to transform the way we live, work and do business? Go to UpTech and let us know


YouTube | LinkedIn | Twitter| Podcast

AI Weapons Detection with Mike Lahiff of ZeroEyes

Big Data Governance and Data Security for a Safer Future with Balaji Ganesan of Privacera