Latent Space 1.0
Empower next-gen designers with Machine Learning
Time
May 2022 - Nov 2022
Tools
Figma, Webflow, Miro, Illustrator, Notion
Team
Gretchen G, founder + product lead
Jiachen D, Market strategist
Kanxuan H, Data analyst
Hanqi S, full-stack engineer
My Role
Co-founder
Product Designer
Deliverables
User interview
User journey mapping
User Flow
Wireframes
High-Fi Prototypes
Usability Testing
Background
In 2022, Gretchen came to me with a well-prepared PowerPoint on Machine Learning aided design. She had a few brilliant ideas on how to empower designers better visualize spatial ideas with ML tools and eventually evolve into an all-in-one product. She wanted me to help her define product strategy and design the product to bring it to life. I love the opportunity to design a product from scratch and since then, we have recruited 3 other teammates to focus on different aspects. This project is ongoing and we are working as a cross-functional team collaboratively.
What you see on this page is my work until November 2022, there’s more to come. We are currently working on launching a public pre-trained model for users to test out. Meanwhile, check out our website here.
Design Question
How could we build an educational tool to help young architects use machine learning without coding experience?
How does machine learning empower designers’ creativity and efficiency?
Solution
A web-based platform with ML tools that helps young designers better visualize spatial ideas and information into speculative drawings and mappings.
Add stops & Customize the trip
01
No-Code Pretrained Models
Discover places and add a stop to your trip for future travels.
Home & Search Page
02
Easy-to-use dashboard
Easily create an image from scratch with our AI-empowered models.
Share models and articles with the community
03
Learn with Community
Create an online community to bridge the gap between architecture and ML
Challenge
01
Limited resources and constraints
This a group of 5 young talents. Since we just graduated out of grad school, we have limited resources and team restraints.
02
Challenge traditional design process
We need to incorporate the traditional design process that has been adopted and well accepted by the architectural community, which started from the Ancient Grace. How to introduce a new way of design becomes our goal.
03
Interdisciplinary workflow
We work as a cross-functional team and thus we need to align among different team members and make sure everyone is on the same page during the product development process.
Cross-functional team alignment
With so many teams and specialties, how do we navigate the project? I created a diagram to visualize the flow of information among different team members.
Research
Research Cycle 1
Explore ML in 2D generation
We looked at ML papers and what models that are typically used in the ML industry. We looked at GAN, DIFFUSION
Research Cycle 2
Deep dive into the industry
We looked at ML papers and what models that are typically used in the ML industry such as GAN, Stable diffusion, etc.
Research Cycle 3
Deep dive into the users
I conducted a user survey to reach out to universities, big design companies, and small practices to check out what designers feel about using ML in architectural design.
Then, we synthesize team meetings to align my research findings with the team. I organized a design workshop, to precisely define pain points and user persona. Since this is a new product, we don’t really foresee how people like working with this product. Our designs need to accommodate many circumstances.
User Groups
We divided the users into 2 types according to their roles and needs. And it enabled us to define their identities as well as all members to be on the same page.
We learned that
01
Most designers are not familiarized with how design works with ML aided tools.
02
People are concerned with the creativity generated by ML models. They want to end up designing their own models once they are familiarized with ML.
03
Most users are not even comfortable with using this technology in design.
Define
Design Principle
HMW help young designers to better visualize their ideas/sketches in informative and speculative drawings?
I want the interface design to be simple, inspiring, and easy to use. Meanwhile, we need to consider maintaining some aspects from traditional design thinking process.
From Traditional arch design
To Machine collaborative design
Define Key features
With the design principles in mind, our team defined 3 major features we would like to develop for the first stage. We categorized the pretrained models to target different steps in designing architecture, which is from site analysis, generating ideas, drawing plans and sections, and generating visual renders.
Ideation
Gretchen and I had several brainstorming workshops in person. Yet most of the time our team work remotly. Figma and Miro allowed us to quickly brainstorm ideas together or asynchronously.
User flow
for 4 Pretrained Models
We did simple wireframe for each of the models to guide the design.
01
Concept Visualization
02.1 (Prioritized)
Site Analysis
02.2
Heatmap Generation
03
Render Optimation
04
Plan Generation
Design
Explorations
Set the Design Framework
With the research in mind, we defined our design goal in the ideation process.
Design Decisions
01
Bar control for non-coder
02
Make visible of the process
Final Design
01
Pretrained Models
Here are demos of the model workflow.
Design your house
Text to image
Create a floor plan
Analyze your neighborhood
02
Website
With time constraints, we used Webflow to build our website to outreach and gather user information. My responsibility is to design the information architecture and visual design of the landing page.
Business Impact
Since this is a new project, our main business goal now is to attract new users, gather and educate users on Machine Learning topics.
1 min
900+
96%
Average view time
Forms submitted
Satisfaction rate
Latent Space 2.0
Our next chapter, we take these ideas and further iterate the features and protoypes. We interviewed, tested, and prototyped a few concepts while working closely with the Machine Learning scientist consultant Sam. We intend to come up with a practical design solution to address the user needs and industry challenge.
