Community Spotlight: AgentHub at HackCamp 2023—Sponsor & Winners

nwPlus
nwPlus
Published in
8 min readDec 1, 2023
AgentHub co-founders (in the back) and their to-be sponsor prize winners (in the front) —working hard!

AgentHub

Max, Rahul

AgentHub co-founders Max and Rahul (left to right) love our mascots—and you should too!

Could you introduce yourselves?

Rahul: Absolutely! I’m Rahul, a co-founder of AgentHub. I did a major in chemical engineering and a minor in software at McGill. I worked at Amazon and then at a cybersecurity firm doing machine learning operations.

Max: I’m Max, also a co-founder of AgentHub. We both went to McGill, where I studied software engineering. I worked at Microsoft in Seattle.

Can you tell us more about AgentHub and its origins?

Max: I left Microsoft to explore building things for the first six months. I was tinkering on random ideas, building free VR apps and exploring different spaces. When ChatGPT came out, I knew that was a big moment. Then AutoGPT, which is like ChatGPT on a loop, is a framework that lets people create autonomous agents. You say, “I want you to solve world hunger,” and it’ll come up with 100 steps and start executing.

I noticed in their Discord that people were having trouble creating an instance because they didn’t know how to use GitHub. So I made a wrapper, a website where you can put in the goal and it runs on its own, and I noticed that people had very niche use cases that they wanted solved. I built a pipeline framework and Rahul hopped on and it’s been heads-down, building ever since.

Do you have a piece of advice for hackers that might want to pursue the same path as you?

Rahul: If you’re interested in building a product or finding a startup, just start building. That’s the number one thing. Everybody who’s here is doing that.

Once you start, I think it’s good to share what you’re building with people. Even if you think it’s bad or filled with bugs, you’ll be surprised at the feedback you get. Make sure you’re always talking to users even from the very early stage. It would be the worst thing to build a completely polished product that you’ve spent two years on, but without actually talking to anybody. Then, as soon as you talk to people about it, you find out that this isn’t a problem worth solving.

Once you’re more experienced with building and you’ve started or shipped lots of products, then it’s even better to validate the idea before you start building. Just talk to people facing the problem you’re trying to solve, pitch your idea for the product, and ask, “Would you use this?”.

At the start, just start building because you’re going to get better and then have a better understanding of what it takes.

Max: You don’t only want to ask people if they would use it, but also how badly they want it. I made this mistake a few times when I was iterating. I built things for months and people were saying that they wanted them, but when it came to paying for it, they were hesitant then. So you have to ask the hard questions and be willing to completely ditch the idea. It feels terrible when you realize what you’re doing isn’t worth doing.

You’ll stumble upon different problems as you’re building. I would never have built AgentHub if it wasn’t for that wrapper around the Agent framework that led me to the other idea. It’s all a domino effect.

What aspects of founding a startup have you found to be the most fulfilling or challenging?

Max: The most challenging part is the start for sure. The very beginning when you have zero validation or traction is all about motivating yourself. One of the best things to do in that case is find a co-founder where you can motivate each other, so you’re not alone while you’re building. In terms of most rewarding, being involved in a community of other builders is great — being able to bounce ideas off other people and make friends who are interested in the same things.

Rahul: It’s rewarding to see people use the product and get value out of it. When you look at the hundreds of people using something I created, they like it, and it’s helping them—that’s something insane. It’s super fulfilling to see that.

Rahul: When you’re working on stuff, there’s so many times where you think, “I don’t know if anybody would use this or anybody would like it — is this important?” It’s really nice to get feedback from people saying that it is [important].

One thing that’s surprising is that the technical stuff is actually one of the easier parts of building a startup. It’s easy to write code and ship a product, but it’s hard to validate the idea, iterate on it, do user interviews, get and incorporate feedback, pay attention to metrics, marketing — all of that is much harder than just writing code.

Why did you decide to sponsor HackCamp?

Rahul: AgentHub is still a new product and we’re in the beta phase. We wanted to interact with users and see what they were using it for, watch them build, and help them if they run into issues. That way, we can improve the product and make it much more usable.

We always went to hackathons during university. So it’s just a nice way to give back now that we’re on the other side.

HackCamp Finalists & AgentHub Sponsored Prize Winners: OOTD

Paul, Carlos, Harry, Abdallah

From left to right: Paul, Carlos, Harry, Abdallah—the team behind OOTD, a smart closet/personal stylist app.

What were the most rewarding and challenging parts of HackCamp for you this weekend?

Paul: For me, the most rewarding and challenging parts came together. When you’re pulling your hair out, trying to get something to work, and then it works.

Abdallah: The most rewarding thing is that we got to connect with the people we never thought we could have, just by the colour of our shirts. We made a really good connection, group project, and good friends as well.

Carlos: Our interests outside of school and humour levels all connected really well. That was something that really helped us: effective collaboration and good team spirit. One of the most challenging things for me was that I’ve never used the backend server. I had no idea of what the concept was to communicate frontend between backend, HTTP, Git, or Post. I had never even used APIs.

Harry: We had a lot of different ideas, technologies, and frameworks that we wanted to use. So the most satisfying part was actually getting everything connected in the end and everything working at the last minute.

You mentioned that you connected based on the colour of your shirts — can you expand on that? How did you become a team?

Carlos: Yesterday in the icebreakers, the prompt was to “find someone who had the same colour shirt”. I already knew Paul, but I found Harry and Abdallah based on the colour of their shirts. And here we are 12 hours later, spending all that time together.

Could you tell us about your project?

Paul: Our project was designed to help people who are stressed when they choose their outfits. We did some research and we found that a lot of people feel this kind of stress — I feel it too. We also kept in mind the theme of accessibility. Our initial motivation was to help colour-blind people specifically. Since colour is a part of choosing outfits, we expanded our research and found that it applied more broadly. We would love to expand the functionality to have it on mobile.

The general idea is that you log on and upload your clothing. It’s like a digital closet, where it stores all your clothing and you can generate an outfit with one click based on that. If you’re stuck, there’s a “get inspired” button.

We managed to use the AgentHub pipeline for our project. You click “get inspired”, and it searches the web with the prompt “latest fashion trend ideas 2023”. Then it takes the top three URLs, scrapes all the content from those websites, and puts it into ChatGPT which provides a summary. That summary goes into ChatGPT again, which provides a prompt to DALL·E, finally returning the brand new outfit idea. It was really cool to use three AI functionalities: searching the web, creating a prompt to ChatGPT, then generating something in DALL·E — all in 12 hours.

What was it like working with the AgentHub co-founders throughout the process?

Abdallah: It was amazing. I spent almost 90% of the time sitting by their side — they helped us a lot.

Paul: I’m very thankful they were here. They were very approachable and willing to mentor and help us. As first-time hackers, that matters; they helped us a lot and made the experience better.

What is one greatest takeaway from this weekend?

Carlos: Even if you don’t exactly know what you’re doing to start, if you put enough energy and time towards it, you can create anything. At the start, I don’t think any of us had much of an idea of how we were going to implement our idea. But here we are, 12 hours later. Thinking back, it was all new to us.

Abdallah: Just start. We took the initiative and we started.

Paul: A hackathon is like a coding marathon. The thing with marathons is that it’s quite painful during it, but that type of big effort often leaves you feeling very fulfilled on the other side. It’s good for people who are exploring their interests to initially have that fuel to dive into the deep end of something.

What’s next for your project?

Carlos: We got AI to generate clothing ideas, but we wanted to use AI to take the clothing within your own closet and pair it together based on what colours go well. The original idea was to help colourblind individuals’ accessibility. As someone who’s colour blind, I struggle sometimes to know what colours go well together. So this is something that I would’ve wanted.

Another functionality we were thinking about adding was a chat integration. For example, “I’m feeling really sad and just want to stay home today,” would give you sweatpants. We also thought about weather integrations, such as shorts if it was hot out or rain gear if it was raining.

Paul: Upon the image upload, we used the Google Vision API to process the image and produce a text description of it. Our idea was that when you upload the images, you have the text descriptions of all the images. Then a user could come on the website and type the prompt for the type of outfit they would want. We’d put it in ChatGPT along with all the data in the database, so you could create a more tailored outfit based on typing, instead of it being randomized.

Find OOTD’s demo here.

Special thanks to these participants for sharing their hackathon experiences. We love hearing from everyone involved in making HackCamp a huge success — we couldn’t do it without you!

Don’t forget to check out the other Community Spotlight from HackCamp 2023: Hackers & Mentors!

Stay tuned to hear more from the nwPlus Community 👋
Find us on social media and subscribe to our newsletter for the latest news!

Written by Jennifer Shui, Content Writer
Edited by Victoria Lim, Marketing Director

--

--

nwPlus
nwPlus
Editor for

The student tech community behind some of the largest hackathons in the Pacific Northwest - empowering thousands of hackers through HackCamp, nwHacks and cmd-f