独立暴富

3.5 Years of R&D vs. Last 30 Days of Funding Chain: How Did We Achieve $3.5 Million ARR with a $0 Marketing Budget During the AI Winter?#

Company: Uizard
Founder: Tony Beltramelli
Monthly Revenue: $292,000

Hello everyone, I am Tony.

Tony Beltramelli created the world's first AI product design tool Uizard. At that time, ChatGPT was still in its infancy, and Tony's product had no market because the concept of AI had not yet become widespread. Later, he leveraged the content that was already ranking high in search engines to ride the wave of AI.

In 2024, when the company's ARR reached $3.5 million, Uizard was successfully acquired. Now, let's hear Tony share his success story.

Table of Contents#

1. The World's First AI Product Design Tool
2. Open Source Viral Spread
3. Building AI Before It Was Cool
4. Finding the Right AI Tech Stack
5. Building with HTML DOM
6. Building User Trust in AI
7. Growth Journey
8. Product-Driven Sales
9. Focusing on Distribution Strategy
10. Where Do We Go Next?

The World's First AI Product Design Tool#

My entrepreneurial journey began with an endless passion for AI and machine learning.

My love for computer science stemmed from my interest in product design. During my graduate studies at the IT University of Copenhagen and ETH Zurich, I was deeply fascinated by the potential of AI and machine learning to empower creativity and solve complex engineering problems.

I co-founded Uizard and served as CEO until the company was acquired.
Uizard is an AI-driven product design tool that simplifies the prototyping and wireframing process for teams. It helps users easily design mobile apps, web apps, websites, and desktop software interfaces.

From founding the company in 2018 to officially launching the product in 2021, we introduced the world's first AI-driven product design tool—long before the current wave of generative AI became familiar to everyone.

Ultimately, Uizard garnered over 3 million users, with an ARR of approximately $3.5 million. At the core of our product, our goal was to democratize design, enabling non-professional designers (such as entrepreneurs, product managers, and developers) to easily achieve rapid iterations and turn ideas into reality.

Today, I serve as the head of AI products at Miro, leading a team to design and develop AI solutions that enhance user collaboration and innovation efficiency. Miro now has over 80 million users and is on its way to reaching $1 billion in revenue.

Open Source Viral Spread#

The idea for Uizard came from my experience as a front-end developer—my first job between my undergraduate and graduate studies. I was shocked by the traditional design-to-development handoff process, which had seen little improvement since the 1990s.

A few years ago, after completing my master's degree, I worked as a data scientist while researching AI in my spare time. I attempted to apply machine learning and computer vision techniques to automate front-end development, hoping to reduce friction and accelerate the iteration speed of product teams. This was in 2017, before large-scale language models (LLMs) or generative AI emerged, and everything was still in an experimental, research-oriented phase, filled with chaos and exploration.

One weekend in 2017, my project suddenly made a breakthrough: I could input screenshots of UI designs into an AI model, and the model would generate the corresponding HTML code. I was incredibly excited, so I wrote a research paper describing the algorithm, recorded a demo video, and open-sourced everything along with the code on GitHub.

The project named pix2code quickly went viral, and many people reached out to me to ask if they could use it in work scenarios. At that moment, I realized that perhaps I could productize this research outcome and create a true SaaS solution.

I had saved enough for six months of living expenses, so I quit my job and fully committed to product development, inviting three smart partners I knew to join as co-founders and began pitching to venture capitalists. Just in the sixth month—my last month where I could still pay rent—we finally completed Pre-Seed funding led by Evan Nisselson of LDV Capital.

Building AI Before It Was Cool#

Building Uizard was a passionate and persistent journey, taking 3.5 years from concept to official launch (from early 2018 to mid-2021). We developed a user-facing SaaS product that included a real-time collaborative canvas and proprietary AI models for generating UI designs, wireframes, prototypes, and code. From proof of concept to a truly chargeable product, we went through countless iterations, user feedback, and technical adjustments to ensure that the technology matched actual needs.

Due to significant R&D investment, we quickly realized that we needed to secure sufficient funding through venture capital to extend the product development cycle and hire a talented team. Ultimately, we completed three rounds of financing, totaling $18.6 million, with the latest round being Series A led by Insight Partners. Before Miro acquired us in 2024, we were preparing for Series B funding.

Funding allowed us to invest heavily in infrastructure, cloud-based GPU training models, and hiring top talent. Our first two employees were a product designer and a computer vision engineer because, in the early days, the four co-founders had to juggle AI engineering, SaaS development, product, infrastructure, accounting, operations, and marketing.

Finding the Right AI Tech Stack#

Our tech stack has evolved over time, but some key technologies have remained unchanged.

• Frontend and Backend: We adopted a Javascript/Typescript tech stack, using React for the frontend and Node.js for the backend.
• AI R&D: We used Python for infrastructure building, model training, and deployment, utilizing frameworks like TensorFlow, Keras, and PyTorch.
• Cloud Services: The entire platform is built on AWS.

When OpenAI released the GPT model, our platform's capabilities were greatly enhanced. We began integrating our proprietary AI models with OpenAI's LLMs and eventually even brought in Anthropic Claude. The introduction of LLMs and generative AI allowed us to chain proprietary models into more complex pipelines, elevating product performance from "good" to "excellent" within weeks.

Building with HTML DOM#

One key challenge that many might not expect is that the infinite canvas in the Uizard collaborative editor is built purely with HTML DOM, not with WebGL or HTML canvas! Yes, you read that right; our entire infinite canvas is composed entirely of HTML elements.

During development, we realized that only by providing users with a canvas where they could freely edit AI-generated content could they manually or with AI further iterate on it.

At this critical juncture, we faced two choices:

1. Build the infinite canvas using HTML canvas and WebGL, like products such as Miro and Figma. While this approach performed excellently, it would require developing extensive software to support drawing, editing, and manipulating UI elements, which could dramatically increase engineering complexity and take a long time.
2. Directly "draw" elements in HTML and insert them into the DOM. Given that we were serving the design market for mobile apps, web apps, websites, and desktop software, and that HTML DOM natively supports UI components like "buttons," "checkboxes," "radio buttons," and "input fields," this approach allowed us to leverage built-in browser functionality to quickly implement editing, manipulation, and interaction features.

We chose the second option because time is precious for startups. This approach enabled us to iterate on the product at the fastest speed and optimize based on user feedback. Although performance issues arose when users built hundreds of UIs within a project, this solution was sufficient to meet 98% of users' needs.

We did our best to optimize React's performance, planning to switch to WebGL after Series B, but ultimately this solution was adequate to support product development before being acquired by Miro in 2024.

Building User Trust in AI#

Establishing user trust in AI is a significant challenge. Especially in 2021, many users were skeptical of AI tools (before the widespread adoption of ChatGPT). Overcoming this barrier required educating users about the capabilities and limitations of the technology.

Another major issue was that we introduced pricing too late, delaying critical product-market validation milestones.

Looking back, I believe we should have involved users early in the development process and focused more on distribution and promotion during product development, rather than just on technology and the product itself. The mindset of "if we build a good product, people will come" does not work. True success lies in combining an excellent product with an efficient distribution strategy (such as content marketing and viral growth).

Growth Journey#

Word of Mouth#

Our growth primarily relied on word of mouth, with 95% of new users coming from spontaneous recommendations and community sharing. Early on, demonstration videos showcasing Uizard's AI capabilities quickly went viral, attracting the first batch of users.

Due to the product's outstanding visual effects (AI-generated UI designs), many early users and customers recorded their usage experience videos and shared them on social media, further driving viral growth.

SEO Optimization#

With the emergence of AI products like ChatGPT, the market demand for AI tools surged. We invested heavily in content marketing early on, optimizing our SEO to rank for long-tail keywords, ensuring our AI-related content had already gained good rankings on Google. When OpenAI launched ChatGPT and the world began to focus on AI solutions, we already had a ready-to-use, high-ranking product, which brought tremendous tailwind effects. In several months of 2023, our monthly revenue even increased by $1 million.

Waitlists#

We frequently leveraged pre-launch waitlists as a growth lever.

The approach was simple: a few months before launching a product or new feature, we would create promotional videos and set up a landing page inviting users to register for the waitlist to gain priority access when the feature went live.

Once users registered, we informed them of their ranking in the queue and provided a mechanism to invite others to register to improve their ranking. This simple gamification design helped us grow the waitlist for Uizard 1.0 from zero to 100,000 users within months.

The same method played a significant role when we launched Autodesigner—the most advanced AI design engine—where over 10,000 users registered for the waitlist every other day.

During 2023 and 2024, we added 100,000 new registered users monthly, sometimes exceeding 240,000 in a single month.

Product-Driven Sales#

Uizard adopted a Freemium model, with paid tiers offering more advanced features and higher usage limits. We continuously adjusted our pricing based on user feedback and competitive analysis, optimizing revenue while remaining friendly to small teams and startups. Regarding pricing strategy, we also referenced and documented some tips.

In addition to a self-service business model, we launched a product-driven sales strategy in early 2023. We identified active users from large enterprises in the self-service model and proactively reached out to discuss deploying Uizard for their entire teams, enhancing their feature usage limits and providing the latest AI upgrades.

This approach helped us successfully sign contracts with several Fortune 500 companies, although early enterprise sales accounted for only 2% of revenue, self-service sales now account for 98% of revenue.

Focusing on Distribution Strategy#

From the beginning, focus on product distribution. Simply building a good product is not enough—you also need a strategy to get users to access and use the product quickly. Initially, focus on a narrow and precise target market and validate your ideas with real users.

Additionally, do not be afraid to set pricing too early; pricing is a key validation metric that can guide product development. Paying users often provide higher quality feedback.

Moreover, the book Hacking Growth (by Sean Ellis and Morgan Brown) has had a profound impact on us. But be careful not to rush into investing too many resources in growth before confirming early product-market fit.

Where Do We Go Next?#

Returning to an employee role has been much more enjoyable than I expected! I am fortunate to work in an organization that gives me enough trust and independence to operate freely.

I look forward to continuing to drive the integration of cutting-edge AI technology with user-friendly applications at Miro, making advanced tools more accessible to the public and helping users quickly turn their ideas into reality. Today, we have over 80 million users, and revenue is on track to reach $1 billion.

Essential Reading for Solo Founders#

1. First-Mover Advantage and Market Education:

   • Tony's early attempts in the AI field (like pix2code) demonstrate that being a first mover sometimes requires educating the market to help users understand and accept new technologies.
   • For entrepreneurs, daring to launch products when the market is not yet mature may face additional challenges, but it can also seize opportunities.

2. The Power of Open Source and Viral Spread:

   • Open-sourcing research results not only garners widespread attention but also accumulates a valuable user base and technical reputation for future productization.
   • This spirit of openness is somewhat counterintuitive, as many believe that commercial products must keep core technologies confidential.

3. Pragmatism in Technology Choices:

   • In the early development phase, choosing the technology that allows for the fastest iteration and validation of the product (like directly using HTML DOM to build the infinite canvas) is more important than pursuing optimal performance.
   • For resource-limited startup teams, "good enough" often has more practical value than "perfect."

4. Balancing Product and Distribution:

   • Relying solely on "a good product" does not guarantee success; early attention should be given to user acquisition and distribution strategies, such as leveraging waitlists, SEO, and word of mouth.
   • This perspective breaks the traditional thinking of "build the product first, then think about the market."

5. The Importance of Early Pricing:

   • Launching a pricing strategy early not only validates market demand but also attracts users who are genuinely willing to pay for the product, leading to higher quality feedback.
   • Delaying pricing may prevent the product from receiving critical market validation signals in a timely manner.

6. The Role of Capital and Team Building:

   • In cases of significant R&D investment and slow iteration, effectively utilizing external funding can buy valuable time for product development while supporting the hiring of key talent.
   • While having versatile team members is crucial in the early stages of a startup, timely expanding the team and bringing in specialists is key to maintaining competitiveness as the product matures.

👉 If you are an independent developer looking to expand overseas and want more similar business analyses and practical cases, please follow our account for the latest and most in-depth independent development business analysis.