Yisha Sun: Deep Learning at Ubiquity6


This is part of a series of blog posts featuring some of the amazing talent we have working on the technology behind Ubiquity6.

This week, we’re excited to be featuring Yisha Sun. Yisha joined Ubiquity6 as a deep learning engineer after six years of grad school, and three advanced degrees in engineering and computer science. She’s dedicated to deep learning research and engineering at Ubiquity6.

Why did you choose to become a deep learning engineer?

I was always very interested in mathematical modeling throughout college and particularly fascinated by mathematical optimization. During my first Master’s program, I participated in a year long exchange to the Tokyo Institute of Technology to study mathematical optimization and it was there that I had my first machine learning class because of the heavy involvement of optimization concepts and algorithms. Deep learning, as a branch of machine learning, also shares the concepts and algorithms that I studied during my Master’s thesis. When I took the deep learning class at UC Davis in 2016, everything felt very familiar and I was impressed by how powerful deep learning had become. I was so excited about deep learning that I began working on deep learning projects at Berkeley National Lab. I feel so lucky and happy to be working on deep learning at Ubiquity6 — it was my dream job!

How did you first learn about Ubiquity6?

I was super excited when I first came across Ubiquity6 on AngelList because it seemed to be the perfect fit for my interests and skill set. I’m interested in techniques that bridge the 2D and 3D world, especially deep learning techniques that can understand the 3D world and render 2D images or other related directions. As someone who has always been interested in visual arts, this was definitely the space I wanted to work in.

Why did you decide to join Ubiquity6?

Ubiquity6 was the best match for my specific interest in deep learning for 2D and 3D reconstruction. I feel so lucky to be taking on what I think is one of the coolest and most challenging problems in the space. Another big reason I chose to join was the people. When I first met the team, specifically Anjney, Ankit and Ryan, I was impressed by how friendly, smart, and motivated they all were. I knew that this was a team that I would love to work with and where I would be excited to go to work every day. After working here for a few months, I feel more and more happy I made the leap!

Can you share some specific examples of your impression of the team?

Sure! I’ve been working closely with Ankit and I’m impressed by his deep technical knowledge in a broad range of areas. I’m amazed by how he manages to work across very different problems and context switch so quickly while making high level strategic decisions with an incredible amount of detail. Another example is one of my great colleagues John — our geometric computer vision expert. I asked him a lot of questions about computer vision when I started at the company and he was just so knowledgeable and so patient and kind. Our projects don’t really overlap but we talk about coding and technical challenges all the time. Our team is very supportive, productive, positive, and collaborative — I’m so proud to be a part of it!

What are you working on at Ubiquity6?

I work on 3D deep learning. Deep learning has been very important in image classification, semantic segmentation, NLP and numerous other applications. Today, there are still a lot of challenges in graphical tasks, and current deep learning architecture design is still limited. However, there has been some research utilizing traditional computer vision techniques. One recent model I’ve been experimenting with is from a paper called Deep Stereo. It can reconstruct an image from images taken at nearby locations and it’s part of a series of research we’re conducting to enable 3D objects and scene recognition, we can get depth prediction in maps, semantic segmentation, classification and more. It’s getting heated!

What is your favorite part of working at Ubiquity6?

I enjoy working with my colleagues, they’re friendly, supportive and very fun, interesting people! I love our fun conversations over lunch and walks to coffee breaks. I also find that I’ve become more productive since working here and it’s amazing seeing all of the cool things that we’re able to build in such a short amount of time. Working on a small team is great because I can see directly how my work affects the product. I also love how our company is so open to sharing ideas and feedback so I’m always learning about other parts of the business. Finally, I always have a lot of fun trying our new features — and while I actually don’t play a lot of games, our game experience is so different from anything I’ve ever experienced and I was totally blown away. I can’t wait to try our next feature and that might be next week because we move so fast here!

What are 3 words to describe Ubiquity6?

Friendly, Fun, Exciting.

Do you have advice for people who might want to join the team?

I think you have to be genuinely interested in the space — come with an open mind and ready to build some cool technology! If you’re excited about what we do here at Ubiquity6, please get in touch!

Source: Deep Learning on Medium