In this episode, we discuss what led Xuchen to embark on an entrepreneurial path after completing his PhD in language and speech processing at Johns Hopkins, his experience transitioning out of academia and co-founding two successful startup companies with a fellow JHU alumnus, and his perspective on intentionally giving back and making a meaningful impact with your career.
Hosted by Michael Wilkinson
To connect with Xuchen and to learn more about his story, visit his page on the PHutures #100AlumniVoices Project website.
In this episode, we discuss what led Xuchen to embark on an entrepreneurial path after completing his PhD in language and speech processing at Johns Hopkins, his experience transitioning out of academia and co-founding two successful startup companies with a fellow JHU alumnus, and his perspective on intentionally giving back and making a meaningful impact with your career.
Hosted by Michael Wilkinson
To connect with Xuchen and to learn more about his story, visit his page on the PHutures #100AlumniVoices Project website.
Michael Wilkinson
Hello everyone I'm co-host Michael Wilkinson and this is the 100 Alumni Voices Podcast, stories that inspire, where we explore the personal professional journeys of a diverse group of 100 doctoral alumni from Johns Hopkins University. Today we're joined by Xuchen Yao. He received his PhD in computer science from Hopkins in 2014 and is currently the CEO of Seasalt.ai. Xuchen, welcome to the podcast. It's great to have you here.
Xuchen Yao
Yeah, yeah, very glad to be here. Thank you, Michael.
Michael Wilkinson
So, you've had quite the interesting career. Soon after you left Hopkins, you founded kitt.AI. You were there for five years, and then you later founded your second company, which is seasalt. What inspired you to go this more entrepreneurial route, like straight from your PhD?
Xuchen Yao
Yeah. So, I guess 1 would be the encouragement of American dream, right? You can just do everything in this land, right? So that's that's one biggest aspiration for me and the other part is really on the plan B side, you know, if everything just tanked, we can always go back to to Google or Facebook to work as a software engineer or researcher, right? So, if you thought, OK, if there's like really really no downside of doing a startup, why not? Why not do it now, right? And also, around the time when I graduated from Hopkins, it was 2014. By the way, I graduated from the Center for Speech and Language Processing, CLSP language and speech processing, right? They were super, super strong in AI. Like currently we are, you know, we are in the next generation of AI which is like headed by all the GPT models from open AI, right. But, nine years ago in 2014, the previous generation AI was super, super hot and strong and indicated by the wide acceptance of success of Amazon Alexa, which was announced and published in November 2013. So, everybody who was working on ARP, which is natural processing. And ASR, which is automatic speech recognition were on a very, very a good track in doing anything awesome with speech and language processing, right? So that's a that's the kind of a background of why we wanted to do embark on this entrepreneur journey.
Michael Wilkinson
OK, so I imagine it's pretty and I know you said well, why not? But I imagine it's still pretty intense to be starting a company right after your PhD. What were some of, like the roadblocks and hurdles that you had to overcome? And where did the kind of confidence to just keep pushing through that come from?
Xuchen Yao
Right. Yeah, so the roadblocks were, as you can probably tell from an accent, you know, a lot of international students had that visa issue. Luckily, we always have to drop back, you know, go back to the times. Luckily that at that time it was the Obama administration and the immigration policy from the US was a kind of friendly, you know, compared to today, right. Today, like a lot of international students just they cannot keep a legal status very easily due to the massive layoffs, etcetera, right? But, at that time it was so friendly that Silicon Valley was also thinking about passing a law, you know, urging the Congress to pass a law. It's something about some startup visa, right, to encourage the international entrepreneurs entrepreneurs to to stay and work for the US right. That thing never passed, but at least at least there were some solutions. There was some hope, right? So, look, I mean, of course, the very first thing we wanted to sort out was like, OK, when I was on F1 visa can I can I just go on and start up a company, stay in the US legally, etcetera. But thanks for the very excellent education from the PhD of program of Johns Hopkins, I was able to embark on this faster track a faster track of the Green card application. We were able to like do the EB1A application in under half a year, you know, because of the academic citation and also international recognition of my PhD work. So, we were able to. I was able to sponsor my own green card without any affiliation, which gave me a lot of freedom to do a startup.
Michael Wilkinson
So, you you talked more about kind of difficulties in maybe staying in the US and all that, but I also just wonder what were the difficulties just in actually making a company from the ground up, right? Like having the confidence to kind of push forward with that.
Xuchen Yao
So that was really about. It was the first time for us to get to know the industry right. Both me and my co-founder, by the way, my co-founder, Guoguo Chen, was a classmate and also graduated from JHU from the same lab as I as I did right. So, there is like a kind of disconnection between academia and industry, especially the startup industry, right? So, in my in my first year I was like really, really thinking about learning about everything we can from a totally different world, right? So originally, I thought I had this very naive thought. OK, how about I just turn my PhD thesis into a startup idea, right. So, I worked on that into three about a three month. I thought, OK, you know, originally, we just write academic papers. Now we just want to turn that into some real-world application so people can use it, right. I had a prototype. It's kind of working and then I talked with one investor that was actually from late Microsoft co-founder Paul Allen's investment group called Working Capital and the investor watched the demo, and he said Ohh yeah, well, good luck with that. You know, you might have, you might be acquired by Google eventually, but I don't quite see any use case for this as how widely it would be adopted. So that 15-minute meeting was this kind of enlightening moment for myself. Like I all of a sudden got epiphany that ohh. OK. Like I'm not only sending this to an academic paper review committee only, not not not to like 3 people to give me good scores, right? But really, to the general public and what the people are gonna gonna buy it eventually, right? So, we'll be talking about this notion of vitamin and the painkiller. Startups got to pay got to sell the painkillers so that people would spend money to buy it while vitamins are just a, you know, good feature, especially during bad times. They do not necessarily become essential for any purchasers, right? So, I thought, OK, we're gonna overhaul everything. Completely dropped my the idea of my PhD thesis and then started something totally new. So that was one awakening moment. You know, I wouldn't say I wasted too much time on that, but the journey was kind of interesting when people have roller coaster ride of up and downs and they have some kind of denial and urencognition from the outside and then finally pulling you out of the the rabbit hole that I just buried myself in right. So, I thought that was super super helpful.
Michael Wilkinson
So, I guess kind of following on that, you know how did Hopkins or how do you feel Hopkins prepared you to go into the space? Because you said you kind of went into it too academically focused, which then you're like, ohh wait, I have to I kind have to switch gears here. So how do you feel, Hopkins actually prepared you to switch gears and to go away in a sense of this, like very academic setting?
Xuchen Yao
So, I think Hopkins helped me in two ways. One, of course, very strong academic background, which helped, which helped my Green card application, right. So, without that, I will never have done kitt.AI, my first startup with my co-founder Guoguo Chen in the first place, right? The second part, is this really, really deep industrial collaboration, especially between my lab, CLSP and Hopkins, right? So, at that time my academic advisors were Ben van Durme and Chris Callison-Burch. Both of them are tenured professors. And they had this very extensive industrial collaboration with Seattle, with Silicon Valley, right. I was able to do some internship due to their introduction. Right? So that's exactly how and why I landed in in Seattle, right. When I first graduated, Microsoft Leader co-founder Paul Allen was just starting the very, very famous AI2 institute. AI2 stands for other Institute for Artificial Intelligence. The short name is like AI and other institute for AI, for artificial intelligence, right, which is they called AI2, right? So, I was doing some internship with the pre-life of AI2, which was Vulcan Inc.’s ARP Group right? And after that I was doing this interview with the AI2 and they said they wanted to hire some research scientist. But I would rather want to do a startup. Right. So actually, my first company kitt.AI became the first company to be incubated in the AI2's incubator program. This they set up special incubator program and that was all due to the introduction of my academic advisor Ben Van Durme and Chris Callison-Burch, their ties in the in the industry.
Michael Wilkinson
That's that's really awesome. So, I wonder when you entered the PhD program, did you know that you wanted to go this industrial startup route? Or was it kind of interactions with your advisors that had these connections that kind of fostered that in you?
Xuchen Yao
Yeah, well, every PhD entering their academic life wanted to be a professor, right? So somehow one way or the other, right? So, but also because, like COSP has this interdisciplinary lab with both the EE and the computer science. We have so much close collaboration at that time with with the Google, with Facebook, with Bloomberg., even with the Snap, right, so a lot of students went on internships in the big tech companies, right? So, I think this is like a very unique and very unique position of standing at the crossroads of computer science, electronic engineering, and AI and a search that gave us there's a potential over there, right. And and in terms of NLP research, JHU has always been the the second biggest in in the US, right after Carnegie Mellon, because we have the LTI, the language or the Language Technology Institute. Right. So, we have a lab. They have institute, but we are the second largest largest and we have been supplying all kinds of students and faculties to the industry, right. I mean not not every PhD would get a professorship. So, majority of them still go back to the industry, right? So, which is you know, which is fantastic.
Michael Wilkinson
Yeah, especially in engineering. I think I was looking it up in mechanical engineering. Which I imagine the numbers would be close to computer science, but I think mechanical engineering something like 68% of PhD students went into the industry over the past 30 years and I think that trend is stealing even more so now, with all the especially like AI machine learning of all their company, our companies wanting to research that more in an academic way to a degree.
Xuchen Yao
Yes, yes, yes, I agree. And then there is also like a good payback back to the school. Right. For instance, my first startup we got, we are the second company to be invested by Amazon's Alexa fund back in 2014 and 2015, right. And then we, Alexa was thinking, OK, shall we sponsor some shall we sponsor some of the academic programs like they could have the Alexis scholarship etc. right? So, they actually consulted us and then JHU was in the first batch of Alexa sponsorship, right. So, the Alexa Fund actually sponsored quite a few PhDs via this vehicle and two JHU. Right. So, we're actually among the top schools. I don't quite recall the other schools, probably Carnegie Mellon and Berkeley, was also in the in the first batch as well, right being, you know, with them and also Stanford with them being in in Pittsburgh or Silicon Valley. Usually they've got the first dibs, but because of this relationship of us, JHU was among the first recipient of Alexa scholarship, and then that helped many, many more PHD's, you know, to finish their degree, right. We think this is a really, really good win-win situation between academia and and and the industry.
Michael Wilkinson
Yeah, Speaking of giving back to Hopkins, I saw that when you exited from ktit.ai, they actually gave a donation or.
Xuchen Yao
Right.
Michael Wilkinson
A donation to the Center for Languages Processing.
Xuchen Yao
Right.
Michael Wilkinson
What led you to do that and what is the importance of kind of like charity and giving back to you?
Xuchen Yao
Right. I guess it's just the appreciation, right? We didn't make too much money from, you know, from the acquisition, but still as a token we thought OK, I mean frankly speaking, me and Guoguo my Co-founder, you know, our our achievements couldn’t have been done without the education from Johns Hopkins, right, and from CLSP. We appreciate it so much. We spent the best four to six years of our 20s at, you know, at Homewood right in Baltimore, right. So, we thought we thought, OK, there's a there's got to be something that we can do as a payback as appreciation. Right. And then what's the, you know what's the best vehicle to do that. You know, of course, a donation specifically to CRSP, right? Because there are so many PhD students, and because we are like, you know, we're in the corner of Baltimore is not really, really well connected to the rest of the US compared to Silicon Valley, right. We want to encourage students to say, OK, you actually have a lot of different choices when upon your graduation and we set up a good example of taking the entrepreneurial journey, right. So, we specifically only sponsored the student seminar and we supplied the seminar food for five years using this donation. So, which is like kind of different than, you know, kind of different than the normal donations, right? You write a check in the school spends it, right.
Michael Wilkinson
Yeah, that's interesting.
Xuchen Yao
So, we put a specific purpose on this donation. Because look, Michael, you are a grad student right now, right? Grad students are always hungry, right?
Michael Wilkinson
That's a fair assessment.
Xuchen Yao
They're always looking for free food. Right. So, seminars going there and then you know you give them pizza, you give them Subs or whatever healthy food you can provide, cookies, right? I think that would bring bringing a lot of joy, you know, encourage people to go to seminars if they say, OK, this is sponsored by a successful startup, right, with a, with a good exit, that would eventually I would say influences some people, right? So that's some kind of real thing. It's not like, hey, Bloomberg just donated like $100 million to found this and that, you know, maybe that has nothing to do much with the the rest of the students, but bringing real food to the grad students and also the faculties we thought we are making a very, very small dent but a deep dent into people's heart impression of this.
Michael Wilkinson
Yep, I I am, I am certainly guilty of targeting the seminars that provide food. Actually, I'm I'm curious kind of on that on that front do you ever do you ever go back and actually speak at some of these seminars that you know you've contributed to to kind of give people a little bit more of a tangible kind of example or inspiration?
Xuchen Yao
Right. We wanted to, but we we never got a chance to go back, you know, due to the pandemic in the past few years, right, and I seriously think that there should be more collaboration between the startup and JHU, especially on teaching entrepreneurship right, and believe school, especially school of engineering, has been having some proposals and collaboration with industry right to to give the students some ideas, right? So, they are good directions to go and the school of the Whiting School of Engineering has been sending people to Seattle and visiting us, and then we set up LinkedIn groups to to get people and alumini together right in a virtual group online, so I think I think we could do much, much more with the students for the school in the future.
Michael Wilkinson
So, I'll backtrack a little bit here and ask a more fundamental question, which is, you know, what does Seasalt do? What does Seasalt.AI do? And how does it relate to your PhD work, since I know you had talked in the beginning about how you describe it a little bit you know. So, yeah, if you could walk me through that.
Xuchen Yao
Right. Yeah, yeah. So just to give some background to the audience, in 2020, my co-founder and I Guoguo Chen who we co-founded kitt.AI with, we started a new company called Seasalt.AI. Seasalt is a short name for Seattle Consulting. Seasalt Seattle consulting dot AI. We still do a lot of natural net processing technologies, right? We saw a special need for the low resource. We we see the under the the the under invested and underrepresented groups with the people who speak what we call the low resource languages. Right. So currently if you think about the general research on natural language processing and speech recognition that has widely been focusing on the major languages such as Spanish and English and sometimes German and French, right, but they're like and also Chinese and Arabics, right. But there are like you know 70% of the world population speak what we call the low resource languages, for instance, like Japanese widely spoken by people in Indonesia on the Java island, right? They you know, that's like probably 200 million people speaking that or or maybe I don't know 50 million people speaking it, but there is not much resource. You know, Google doesn't work very well and then you know ChatGPT doesn't work in that language pretty well. But they're still like real human beings who speak of who speak of that right. So, there's, like, no Amazon Alexa or or Apple Siri that speak that that, that language. And those underepresentative groups do not have those AI tools to help them in their daily life and in their business transactions, right? Even for English in Africa, you know we have the Nigerian English or the Pigeon language, right, which is a very, very remote dialect of English. And that was not very taken care of, right? So, we think there is a great great opportunity to invest in this under representative under representative people groups and language group, right, that's where where Seasalt.AI is gonna be in terms of bring some welfare and then all some products to them.
Michael Wilkinson
And are you working? Is your company working with like the governments of these various countries? Is it working just like businesses within these various countries? You’re mainly working to kind of get this out to the larger populations?
Xuchen Yao
Right, yeah. So mainly working with the big enterprise companies, which has a lot of consumer customers, right. So, we think that's going to be the biggest multiplier, right? If you're working with, for instance, the biggest travel companies in in Taiwan or the biggest banks in Indonesia or in the Philippines. So, by using our technology, they could apply those in their consumer customers, right? So, our mission Seasalt.AI’s mission is called self B2B with a B2C in mind. B2B stands for business to business and B2C stands for business to consumers, right. So, whenever we want to solve a problem of a business, we think about that their customers which their consumer customers right. So that's a that's we want to do this, you know, second degree consideration by providing the best consulting and products to the businesses so that the consumer customers would have a direct impact, right? So that's our goal.
Michael Wilkinson
And to understand a little bit of the behind the scenes, is it is the method to make models that are just more universal or you just to make models for the different languages that require it just because it's like it's not generalizable in that same way?
Xuchen Yao
Right, right. Yeah. So, one year ago I would say OK, that would be like making individual models for each country and for each language. But now you know, ever since OpenAI released GPT 3.5 and the ChatGPT since last November 2022, right and now we're all talking about the foundation model. So those are just truly amazing language models that can do all kinds of different things. It has implemented and captured the idea of Noah Chomsky’s grand notion of universal grammar, meaning that everybody on earth are just speaking the same language, using different syntax and grammar, right? And that's what we call LLMs, large language models has truly embedded and included all the aspects of all kinds of different languages. Now you can use the same language model to do different tasks in different languages. Of course, English, Russian and Chinese are still the best languages and Spanish is supported by them, right? But you know, we can essentially use the same model to work on 100 different other languages. So that's actually our goals. To answer your question, it's going to be making a general model widely applicable to a lot of other populations and other languages.
Michael Wilkinson
Right. Which is so interesting because you know you mentioned like the main ones have support are English, Russian, Spanish, Chinese and those are all so different from one another. So, I I think this kind of the idea that you could go universal makes sense if you can have models that can handle those four. Presumably you can have models that can handle more than those four different combinations of that.
Xuchen Yao
Right, exactly.
Michael Wilkinson
You talk with a lot of passion about the work you do. And I'm interested in, you know what are some of the fun aspects of the stuff you do or like what are some of the fun projects you're working on or just like the ways that the job has been fun for you?
Xuchen Yao
I see a lot of cross-border transactions. I thought that's kind of that's kind of fun, right? I see the the world is gonna go is going to be even more global than ever, right? So, a lot of our customers are, what do we call cross-border companies who do business in a different country. Right. And then there's like so much collision of culture shock and how people do different businesses in each in each country, in each world, right. So, we thought, OK, that's kind of that's kind of fun, right? Like if you think about this notion of global village as the with other human beings in the world, it's really not about, you know, setting to a very unique market on a very unique language. It's really about, OK, how to facilitate the transactions between different nations, between different speakers of different languages, right? So that's the kind of very, very interesting by putting ourselves because of our unique international background we do some businesses in the US and for Southeast Asia with, you know, while not speaking in the native languages right. So, we see a lot of interactions between people with different cultural background etc. So, I think that's kind of fun.
Michael Wilkinson
On a kind of fun funny note. So how does that affect your sleep schedule cause I imagine being on Seattle time interacting with people
Xuchen Yao
Right.
Michael Wilkinson
Across the world, you have some meetings at some very weird times.
Xuchen Yao
Yeah, it is. It is very, very awkward times. I just came back from a trip from Singapore from Singapore to Seattle. That's a 17-hour flight and when you fly over there and you know you fly, you take a flight today and they are also 13 hours ahead of us, right? And I'm sorry, 15 hours ahead of us plus the 17-hour flight. So, I'm gonna. I'm gonna be there the day after tomorrow. Right. So, you take the flight on Monday and you're gonna be there on Wednesday and you have no idea where where Tuesday was, right? So that's a that's a kind of super, super interesting when you are in a different country
Michael Wilkinson
Well, and and you get the inverse coming back right?
Xuchen Yao
The inverse coming back is slightly better because we are counter setting the time difference, right? So, I thought, wow, OK, yeah. And also, my sleeping schedule also got two small kids two years old and eight months old.
Michael Wilkinson
Congratulations.
Xuchen Yao
Well, thank you very much. They they have been they have been getting getting they have been messing around with my sleeping schedule anyways.
Michael Wilkinson
So, I I I guess on that point you know, I imagine maybe managing the second company might have some more ease in managing the first, but how do you how does your kind of work-life balance come into play when you have such an important role in this company? And now you also have two small kids that you have to manage and all this other stuff.
Xuchen Yao
Yeah, there's really no work life balance, to be totally honest. Yeah, like doing the second company is already difficult. It's not easier than the first one, right? Like they always say all the great things have to be invented twice. We are a serial entrepreneur, me and my co-founder, right. Now it's not, you know, easier than the first time ever, because every time you do it, it's like there's always something, something different. Right. So, I hate to do it all over again, which means that I want the second company to be really, really good. I don't want to do a third company anymore, so that's that's what I see here, right. And also, we got two pandemic babies, right? Like 2 years old and an 8-month-old right. By handling the two babies alone would be super difficult already right? At this point I really, really want to thank my wife and my parents and my in-laws, right. So, they’ve been a great help during the pandemic day and night, right? Literally, it's so difficult to do this and I cannot say that I got any life balance at all, you know.
Michael Wilkinson
Yeah, it's like my my wife and I worry about, like, kids during postdocs. I can't even imagine kids during a CEO.
Xuchen Yao
Right. Right, right. Yeah. Yeah. And also like in the 20 employees, right? And now during the time of March 2023, every big tech company is laying off right? We were able to keep afloat. We didn't lay off anybody.
Michael Wilkinson
That’s wonderful.
Xuchen Yao
Right. So, because I know if if if I let any of the employees go, they're gonna know where to go, right. And, you know, some some people got mortgages right. So, there's, like a great, great responsibility to the company, to my colleagues, and also to my family. So, like everyone just gotta gotta work super, super hard to keep everything float, yeah.
Michael Wilkinson
So, we're getting close to the end of time. So, I'll ask kind of two final questions to wrap up, the first of which being you know, what advice would you give to the current PhD students or those who are close to being finished as far as like wanting to pursue something in like an entrepreneurship or just pursue these goals that are not as standard in in, in PhD life?
Xuchen Yao
Yeah, there is so much I didn't know as a PhD, so I would always go back to Steve Jobs quote which is stay, stay hungry and stay foolish, right? So, we always say that, you know when we hire, we hire for attitudes and we rank attitudes better than your skills and better than your knowledge. For PhD students, they have a lot of knowledge, and sometimes they have a quirky attitude, but eventually what's going to help you a lot would be the right attitude sometimes with the right skills, but it doesn't matter if you do not have that knowledge or the right skills. Once you have the right attitude, you can learn everything, right? So that's that's what we value the most right? Just go outside, try everything and stay hungry with humble, humble attitude towards anything. And then you'll have a lot of success.
Michael Wilkinson
That's, I think, a wonderful piece of advice, and I think a lot of students sighed a sigh of relief when you said you don't need to have the knowledge necessarily. So, I'll I'll thank you so much for, you know, taking the time to do this, I'll end off with this. What inspires you right now, in things you're doing, in your life, whatever that may be. What is a source of inspiration for you?
Xuchen Yao
It's really about making some, making some small dent in the, I mean, making some small dent and then not to waste my life, right. Like we had so much education from kindergarten all the way to PhD. By the time I graduated, I was already 30 years old, right? Like you had to do something to pay back to the society and do something fun. Have that wow moment to bring it to your customers. Right. So, I think that's what everything goes, what keeps you going, right? From within my company, I always tell my colleagues like when you were younger, you're thinking about what kind of skills you can put on your resume, what kind of achievement, right? You know, some people can type like 600 words in one minute and they would say, OK, I'm a great quick typer, I can use Microsoft Office. I can use the Salesforce the CRM. It's actually I have this certification that certification, but when you grow older, 20 or 40s or 50s, you think about what do you want to put on a tombstone? I wouldn't put OK Good Excel user, office you know type quick type on my tombstone, right? That's not achievement for life. That's not what you're remembered for, right? So, you gotta think about, OK, what you can leave to the rest of the world when you when, when, when it is your time. Right. So that's the kind of inspiration.
Michael Wilkinson
Well, yeah, I think that is a wonderful sentiment to end on and and definitely an inspirational goal. Thank you so much for taking the time to do this. I look forward to following up and speaking with you further, but it's really been truly wonderful.
Xuchen Yao
Yeah. Yeah, thank you as well. I'm very honored to be part of the JHU’s Alumni right? Yeah. You know, go Hopkins, go Blue Jays and everyone, thank you very much, Michael.