This founder left academia, built four failed products, then hit $12.5k/mo.

Selene Lee quit academia to build a portfolio of projects. The first four products flopped, but product #5 — SocialCrawl— is at $12.5k/mo and counting.

Here's Selene on how she's doing it. 👇

Leaving academia

I don’t have a traditional tech background. I spent nearly nine years in academia and completed my PhD in neuroscience at Oxford in 2023. My research focused on how the human brain ages and the biological mechanisms underlying dementia.

During my PhD, I met my cofounder, Oscar Lee, who was pursuing a PhD in AI and physical computing. In 2024, we founded Ridio and began building side projects alongside our full-time research jobs. I left academia in the summer of 2025 to work on the company full-time. Over two and a half years, we built four AI consumer products based on our interests and expertise, but none gained enough traction to become viable businesses. Our fifth project, SocialCrawl, is making about $12.5k/mo.

SocialCrawl gives developers access to live, clean public web data through a single API. It began as a social media data scraper for AI agents. Something like Plaid for platform data, but it has since expanded into news and e-commerce. We now support more than 44 platforms and 325 endpoints. Every platform uses the same standardized data format, API key, and billing system, making integrations much easier for developers. We’ve also built products such as Universal Platform Search and PRISM, which enable cross-platform searches and AI-enriched results. Our goal is to become the most developer-friendly way for enterprises and AI labs to access reliable, low-latency public web data.

Finding the idea

I loved research because I always learned and tried to answer difficult questions. But after spending years at a laboratory bench, I realized I wanted something more practical and dynamic. Research can move slowly. You might spend months running experiments before seeing the outcome. I wanted to build something people could use immediately and feel my work had a more direct effect on their lives.

I also wanted more autonomy and a chance to create something of my own. Oscar and I started building side projects alongside our research jobs. By the time we started SocialCrawl, I had left academia to work on the business full-time, and we remained bootstrapped. One of the biggest lessons from our earlier SaaS products was building things we didn't use ourselves. This time, we wanted to begin with a problem we experienced daily.

At the time, I was managing social media accounts generating around two million views a month. Keeping that momentum meant spending hours daily researching content and hunting for opportunities. Oscar had built automation systems during his time in physics labs, so he created a tool to help me. We soon realized we had plenty of data, but accessing it was a mess. Fragmented APIs and separate subscriptions spread it, with each source returning it in a different format. That internal tool eventually became SocialCrawl.

Earlier this year, Nathan joined our team. Based in South Korea, he had previously worked as an AI transformation consultant. He had seen companies encounter the same problem: they wanted to build products around external data but couldn't rely on its infrastructure. His experience confirmed that other businesses dealt with the same frustration.

We kept market validation simple. Were people already paying for a solution? Yes. Existing competitors proved demand, while our conversations with users showed us where those products fell short. The same complaints repeatedly surfaced: unreliable data and limited coverage. This became our focus. We believed we could build a more dependable product with broader coverage, giving customers a strong reason to switch.

A two-week MVP

We built and launched the first version of SocialCrawl in two weeks. We followed a simple rule: ship quickly, then iterate. Oscar built it himself using Claude Code, which sped up the development process. We kept the first version deliberately small to put it in front of users and start learning from them as soon as possible.

We made most of the early decisions quickly. We chose the name in about three minutes and created the first logo in fifteen. After spending a few hours studying competitors, we chose a core feature that customers were already paying for elsewhere. That gave us a sensible starting point without overbuilding. Once the product was live, we could learn what users disliked about existing services and decide where SocialCrawl could be different.

The MVP went live at the end of those two weeks, and we got our first paying customer within 48 hours. We stayed close to our early users and added features based on the problems they described. That feedback gradually pushed us beyond the features competitors already offered and toward proprietary endpoints of our own, including PRISM and Universal Search.

Because Oscar built the product himself, the initial version cost very little. We didn’t have a large budget; most of our spending covered a Claude subscription, the domain, and the third-party APIs required to run the service. We covered those costs ourselves.

The stack

SocialCrawl is a Turborepo-based monorepo written in TypeScript on Next.js with Hono API. We use Vercel AI SDK and Gateway for AI models, including OpenAI (gpt-5.4-nano/mini, gpt-5.5), Anthropic (Claude Haiku), Google (Gemini Flash + embeddings), xAI Grok, Perplexity Sonar, Cohere rerank, and Azure gpt-image-2 for images.

We deploy the codebase on Vercel, using Cloudflare for domains and R2 storage. We use a Postgres NeonDB with Drizzle for the database, Upstash Redis and QStash for caching and background jobs, Better Auth for authentication, and Stripe for processing payments. We are currently on Max 20X plans for two Claude Code subscriptions, a 5X plan on Codex, and SuperGrok Heavy.

Our typical workflow involves Claude (Fable) for planning and code reviews, Grok 4.5 for implementation, and Codex/ChatGPT for more general work like admin, codebase optimization, and SEO.

Upgrades and retention

SocialCrawl has a usage-based model with four tiers: Free, Starter at £15, Growth at £49, and Pro at £299. Credits never expire, and we require no subscription. New users receive 100 free credits without entering card details, allowing them to test the API immediately. We charged from day one because willingness to pay provided a much stronger signal than sign-ups alone.

The free tier caused problems at first. Because it didn’t require a card, hundreds of bots created accounts to collect free credits. We added safeguards to control the abuse but decided to keep the plan. Developers often want to test a few real requests before committing to a subscription, and free credits make that possible.

Revenue accelerated when Growth customers upgraded to Pro. Once SocialCrawl becomes part of a customer’s workflow, their usage increases, and they need more API credits. That creates a natural path from a £49 customer to a £299 customer.

Some companies need more than the Pro plan provides, so we also offer custom enterprise plans. Nathan’s relationships with Korean companies helped us initiate our first enterprise conversations and partnerships. We are onboarding more global firms heavily using social and e-commerce data. We have become their reliable source of data and expect a long-term partnership.

Retention has mattered as much as upgrades. We speak regularly with users about the data they can’t get reliably from other providers. As the customer base grew, we saw the same requests for particular platforms and endpoints. We built the most common requests, and some eventually became proprietary products such as PRISM and Universal Search. Those additions made SocialCrawl harder to replace and helped customers get more value from the same service.

The biggest lesson I’d give other founders is to stop treating a low price as their main advantage. With our earlier projects, I tried to reduce the price because I assumed the cheapest option would win. It didn’t. Customers will pay a fair price when a product reliably solves a painful problem. I wish I’d focused on making that value clear much sooner.

Give first, ask later

Our previous products taught us that distribution matters as much as the product itself. Before building SocialCrawl, I grew an audience around my work. I now have around 2,200 followers on X, more than 5,000 on Threads, and nearly 800 YouTube subscribers. Those numbers are not extraordinary, but they helped me reach the right people. SocialCrawl is a developer tool, so I focused on places where developers and indie founders already spend time.

On X, I build in public almost every day. I share our work, product updates, and the problems we're solving. By launch day, people had watched SocialCrawl take shape and knew its creator. That familiarity helped us earn trust and attract our first users.

WhatsApp communities were our strongest launch channel. I joined several groups for indie founders long before SocialCrawl launched, participating as a genuine member. I signed up for other founders’ waitlists, tried their products, and gave them feedback. When I shared SocialCrawl in those groups, they brought us hundreds of sign-ups on launch day. The relationships came first, making a big difference.

We also fully used our existing networks. When SocialCrawl launched, I told everyone I knew: friends, LinkedIn contacts, former colleagues, and people in WhatsApp communities. Some became users; others opened doors we couldn’t have reached through social media alone. Nathan, for example, had relationships with several large global companies in South Korea. We quickly put together a deck and booked a meeting through one of those connections. That meeting developed into a partnership. Your first major opportunity may already be in your network.

My general marketing approach is to give people something useful before asking for anything. I also run a personal blog, Friday AI Club, where I share AI workflows and explain the tools I use. It hasn’t been a major source of sign-ups so far. Blogs take time to grow, but I see it as a longer-term way to earn trust by sharing useful knowledge without attaching a pitch.

My advice is to find where your users already spend time and start showing up before you launch. A small community of relevant people can bring better early traction than a viral post seen by people who will never use your product. Give first, ask later. That was my rule when launching SocialCrawl, and it still shapes how I market it today.

Build your audience early

The hardest part was getting people to care. AI makes building products much faster, shifting the bottleneck to distribution and finding paying customers. So much online noise exists that earning someone’s attention can feel harder than building the product itself.

With our previous projects, I didn’t understand our ideal customers well enough. I chased virality across too many platforms and wasted both time and money. With SocialCrawl, I had a much clearer idea of who needed the product and where those people spent their time. I then needed to find the right way to reach them and stay consistent long enough for it to work.

I started my X account with three followers. Even after reaching around 500, many of my posts still received fewer than 100 impressions. It often felt like I was shouting at a wall. Still, I posted every day and left 40 to 50 thoughtful replies on other creators’ posts. Staying consistent when nothing seemed to be happening was the hardest part. Eventually, some posts reached much larger audiences and started sending organic traffic to SocialCrawl. My account has now grown to more than 2,200 followers. I followed the same approach on Threads, YouTube, and LinkedIn, focusing on platforms where our potential users already spent time.

If I had to start again, I’d begin building an audience long before launching the product. Posting every day felt uncomfortable at first. I didn’t know what to write, and I hated speaking in front of a camera. I wish I’d understood sooner that I needed to be bad at it before becoming comfortable. Posting still feels cringe sometimes, but those awkward early posts eventually opened doors and led to relationships I wouldn’t have found otherwise.

The value of founder meetups

One of the most helpful habits I’ve developed is regularly speaking with other indie hackers, founders, and builders. I use Luma to find nearby founder meetups, hackathons, and seminars. Meeting people in person has been especially useful because the conversations tend to be more open and honest than what you see online.

Those conversations have shown me that most founders struggle with many of the same problems. I’ve often met someone who has already dealt with an issue I’m currently facing, and their experience has helped me avoid mistakes or approach it differently. It’s also motivating to hear what other people are building. Whenever I feel stuck, spending time around other builders usually gives me fresh ideas and helps me get moving again.

Don't hide inside product work

Here's my advice: Don’t hide inside product work. With our failed projects, I spent months building, iterating, and adding features. I knew marketing mattered, but I kept postponing it because it felt hard and uncomfortable. Building was the part I enjoyed, so it was easy to convince myself that another feature was the best use of my time.

Start marketing before the product feels ready. Speak to potential users every week and choose one distribution channel where you can show up consistently. Treat marketing as part of building the business, not something you’ll get around to after launch. If the right people never hear about your product, another feature won’t make much difference.

What's next?

Our long-term goal is to make SocialCrawl the most convenient and reliable data layer for AI agents and the people building with them. Our ambition goes beyond proxy infrastructure. We want customers to have one dependable way to access public web data without maintaining dozens of separate integrations. That convenience and reliability are already a big part of why customers stay after switching.

In the near term, we’re onboarding larger enterprise customers and expanding the number of platforms and endpoints we support. We currently cover more than 44 platforms and 325 endpoints, and we want to keep increasing both the breadth and depth of that coverage. Partnerships will help us reach more platforms and understand what larger customers need from the product.

We also plan to explore external funding. We’ve bootstrapped SocialCrawl so far, but additional capital could help us expand coverage faster and strengthen the infrastructure as enterprise usage grows.

One of my personal goals is to build a stronger personal brand by talking honestly about both the failures and successes behind the business. I’m still working out exactly what my positioning should be, but I plan to keep growing my presence on X, Threads, and YouTube.

And further into the future, I’d love to reconnect with my neuroscience background. I still care deeply about the field, and Oscar’s background in AI gives us an interesting overlap. One idea we’d like to explore is an AI system inspired by the architecture of the human brain to make computing more energy-efficient. It’s still an early idea, but it’s a direction I’d be excited to pursue.

You can follow along on X, LinkedIn, Threads, and YouTube. Or check out Friday AI Club and SocialCrawl.