I Got Rejected By 22-Year old CEO
Chapters7
Describes how AI relies on massive labeled data and how a scalable data platform is essential for safe and reliable models.
Joma Tech riffs on his early Scale API interview experience, the startup’s data-labeling pivot, and what it teaches about rejection, persistence, and building for AI.
Summary
In a candid retrospective, Joma Tech walks us through his 2016 encounter with Scale API (now focused on labeling for AI training) and the quirky path of his career at the time. He recalls applying as a first engineer for Lucy Guo and Alexander Wang, sharing a take-home project that simulated how contractors would annotate images with bounding boxes. The clip includes a live demo of his UI for drawing boxes on images, handling multiple requests, and sorting by urgency—demonstrating his hands-on approach and the kind of work early startups prized. Joma notes Scale’s evolution from a broader “data labeling for trading data” platform to a refined labeling-for-model-training focus, a pivot that mirrors the realities of AI data pipelines. He peppered the tale with personal anecdotes—matching with Guo on Tinder, being ghosted after the interview, and the self-deprecating humor about his front-end experience (mean stack, Angular, and the self-admitted gaps). The video weaves in a glimpse of the startup culture Venn diagram: doing many hats, rapid feedback loops, and the immediacy of real-world feedback from founders. While the main thread is a humorous confession about rejection, the core takeaway is about learning from early-stage interviews and the value of building practical demos to showcase your fit for a small, fast-moving team.
Key Takeaways
- Scale API started as a broader API for human labor before narrowing its focus to labeling for model training, aligning with how AI pipelines rely on human-annotated data.
- Joma Tech’s take-home project demonstrated how contractors would annotate images by drawing boxes, including handling multiple images, urgency sorting, and saving results to a database.
- Lucy Guo (co-founder) and Alexander Wang were the early interviewers who evaluated him for an engineer role, highlighting the ‘wear many hats’ expectation in startups.
- Joma’s personal anecdote about being ghosted after the interview underscores the ambiguous nature of early startup recruiting and the persistence required.
- The video showcases a practical, hands-on UI build (mean stack choices and a lack of deep frontend focus) as a testament to the kind of work startups valued at the time.
- The exchange includes a light self-deprecating humor about dating app connections and the candid confession of not fully embracing frontend frameworks, illustrating candid founder-founder and founder-employee dynamics.
Who Is This For?
Essential viewing for aspiring startup engineers and product folks who want a behind-the-scenes look at early-scale data labeling in AI, plus practical demos that demonstrate code-and-passion over polished slides.
Notable Quotes
""This company used to be called Scale API, and it wasn't focused only on labeling trading data. They built an API for human labor.""
—Describes Scale API’s origin and pivot toward labeling for model training.
""So I applied. Here's the email exchange we had. Lucy Guo... I actually matched with her on Tinder a year ago. She didn't respond to me, but but then again, my opening message wasn't that great.""
—Highlights personal anecdotes and the startup’s informal hiring culture.
""The project asked me to create post endpoints for them so their customers can send annotation box requests... draw boxes around whatever they want and then send them back to the customers""
—Shows the actual take-home task and how the workflow was framed.
""They invited me for an interview and Alexander the CEO interviewed me. We just chatted and then he asked me an algorithm question... a fine median in a data stream question.""
—Recounts the interview experience and the kind of technical questions asked.
""I’m ghosted after the interview... I would have been a trash employee. What's popping, guys?""
—Candid self-deprecating moment illustrating startup dating-recruiting reality.
Questions This Video Answers
- How did Scale API evolve from labeling for trading data to AI model training labels?
- What does a first engineer interview look like at a tiny startup like Scale API in 2016?
- Why do startups ghost candidates after interviews and what can you learn from it?
- What is the mean stack and why did Joma use it for Scale API's early UI?
- What does it mean to wear many hats in a startup, and how does that influence hiring decisions?
Joma TechScale APILucy GuoAlexander WangAI data labelingdata labeling for AImean stackAngularfrontend toolsstartup hiring stories
Full Transcript
What are you doing for Whimo, Cruz and Uber? Yeah, so what we've done at scale is built the data platform for AI. So AI is really built on top of data and these algorithms require billions and billions of examples of labeled data to be able to perform in a safe or reliable way. So this 22year-old interviewed me for a full-time position and rejected me. Just kidding. He was 19 when he interviewed me and rejected me. That was 3 years ago and I was a senior at Waterlue. Well, in my defense, they never actually explicitly rejected me.
They just ghosted me after the interview. I'll show you the emails later. Let's take a look at their old website. This company used to be called Scale API, and it wasn't focused only on labeling trading data. They built an API for human labor. We're going to go back to August 2016. Huh. I don't think that's it. Ah, okay. That's more like it. So, kind of like Task Rabbit, but with uh without the in-person stuff. So, like phone surveys, transcription, e-commerce tagging, some categorization, outsourcing [ __ ] work essentially. Now they only focus on categorization aka labeling for model training.
So refined [ __ ] work. Scale was founded by Lucy Guo and Alexander Wang. Man, even his name looks like a startup. Anyways, back then there were only two co-founders and they were looking to hire their first engineer. So I applied. Here's the email exchange we had. Lucy Guo. I actually matched with her on Tinder a year ago. She didn't respond to me, but but then again, my opening message wasn't that great. So, apparently my resume stood out cuz I did a lot of different roles. So, that's good. As a startup, we look for someone who can wear many hats.
Well, that's good. They quickly hit me up with a take-home project. Oh [ __ ] Well, I mean it's too late now. The project asked me to create post endpoints for them so their customers can send annotation box requests aka send images and then we or scale API will have their contractors draw boxes around whatever they want and then send them back to the customers where they will probably use it for some machine learning model or something. So, this project just focuses on this part and the UI for the scale contractors to actually draw the boxes. Actually, I think I still have the code for this.
Let me whip it out. All right. So, here's what I built so that the contractors can use to draw boxes. All right. Let's send it some requests. So, here you see there's four images or four requests I'm sending. All right. And then we'll see this. And here you go. So, you could see now that you have a few requests. And here you see a picture of um a few cows. And here it says you have to annotate baby cow and big cow. And then the instructions it says draw a box around each baby cow and big cow.
So that's cool. So I think it's pretty simple. You just, you know, draw them. Baby cow, baby cow, baby cow. Oh no, that's not a baby cow. So you know you can exit out switch to big cow and then annotate the big cow. Then you can also uh reset it. You could also say it's broken. I don't know if ever it's broken. And then there's also the urgency here. You could also sort by urgency if you want. So here you go. You have urgency. Sort by urgency. Sort by date. You could also click another one if you don't want to do this one right now.
Uh yeah, I think that's pretty much it. So let me just annotate this. Baby cow, baby cow, baby cow, and then mama cow. This will be saved in the database and sent to the customer. And the customer will receive will receive these things. And then you just submit. And then you submit it. And then here's another example. Um I think this is a picture of a K-pop group called No. Okay. I I I know it's Blackpink. Okay. I know it's Blackpink. So here it says draw a box around the best Blackpink member and then the objects it's like my bias.
Okay. Um let's see. Okay. Um okay. Okay. And then Oh. Oh [ __ ] Are you serious? All right. So, I think you get the point. And um they also told me to use the mean stack um if I can, but to be honest, I wasn't much of a front-end guy. So, I ended up using the men stack, you know, very patriarchal of me. Um, I had some Angular, but it was clear that I wasn't really using the framework, right? But, I mean, they said they loved it, though. Uh, but I have a feeling they didn't look at the source code until later on when they ghosted me.
So, no hard feelings, though. I mean, I I' I'd ghost myself, too. I would have been a trash employee. What's popping, guys? Here's a day in the life of an early startup employee. So, here's me coding right now. So, they invited me for an interview and Alexander the CEO interviewed me. We just chatted and then he asked me an algorithm question. The question was a fine median in a data stream question. He had to give me a hint for me to solve it. So, um, that probably didn't help my chances. And, uh, unfortunately, he didn't respond to me after the interview.
And then, as you can see, here's my last desperate attempt after a month. Oh, no. Actually, actually, that was my last attempt.
More from Joma Tech
Get daily recaps from
Joma Tech
AI-powered summaries delivered to your inbox. Save hours every week while staying fully informed.



