This full-time data scientist built a $7k/mo portfolio of side projects

Ramsri Goutham Golla is a full-time data scientist with a portfolio of side projects. That portfolio — which includes Questgen.ai, Supermeme.ai, and AiArtist.io — has over 2.5M registered users and is bringing in nearly $7k MRR.

Here's Ramsri on how he's doing it. 👇

Building three apps while working full-time

I am Ramsri, a microSaaS builder from India, currently building three apps alongside my full-time job as a data scientist:

• Questgen.ai is an AI quiz generator that takes text, PDFs, or URLs and generates quizzes such as MCQs, True/False, and Fill-in-the-blanks. Since its launch 4 years ago, it has generated over $150K USD in revenue and has over 150K registered users.

• Supermeme.ai is an AI meme generator. It takes any text and generates complex, multi-panel memes ready for social media posting. It has over 2 million registered users and achieved Product Hunt #1 of the day upon its launch.

• AiArtist.io is an AI motion graphics generator that creates videos from any text. Recently launched, it has over 20 paying users.

At their peak in 2025, the apps hit $100k in total ARR. Currently, they hover between $6k and $7k MRR, with more than 2.5 million registered users.

A bet that paid off

In 2021, I explored ideas at the intersection of AI and education. As AI emerged, I saw quiz generation as an ideal challenge, believing technology would significantly improve its quality in the coming years. That bet paid off; after 2023, text-to-quiz generation gained momentum.

I spoke with several educators to understand their quiz generation pain points, and learned that they outsourced quiz generation to item bank companies. I also found out that they conducted several stages of review. From these conversations, I saw an opportunity that would help companies manage quizzes with their internal workforce instead of outsourcing.

That led to Questgen.

I first used no-code tools like Bubble to create a working version. But the functionality had limitations because those tools abstracted code, so I learned a full-stack platform (NextJS) and built a full-scale app.

The stack is currently NextJS + Supabase for the backend and authentication. And nowadays, I use tools like Claude Code to vibe code and add features.

Converting to a subscription model

Initially, we launched as pay-per-use, but we soon converted to a subscription-based model. Plans start at $10-$15 for the starter plan and go up to $50 for higher usage. The annual plan is 40% cheaper than the monthly plan.

We have 60-70% margins after covering LLM costs.

Open source and SEO

I initially open-sourced Questgen as a GitHub repo. Anyone could download the models and self-host quiz generation locally or within their organizations.

Open source was a great way to promote and launch. People are more likely to share open-source projects, so it quickly gained traction from word of mouth. Beyond that, GitHub drove traffic, and I wrote a lot of blog posts on the general topic of question generation. That kickstarted SEO.

After Questgen gained considerable monthly traffic, I added its web app URL to the GitHub repo's main page (README). This allowed people who didn't want to self-host to become customers.

Since launch, our focus has been on SEO for both products. That includes alternate landing pages, keyword targeting, and programmatic SEO pages for individual memes.

I also love building in public. From the start of an idea, I always share my progress online, primarily on X and LinkedIn. I post almost every day. It's also important to launch and relaunch!

A viral PH launch

For Supermeme, we launched on Product Hunt. It went viral and ranked #1 for the day. Here's how:

1. Since none of us had a huge audience and we hadn't done a PH launch before, we reached out to Hunters on X until we found one who was interested in our product. They then launched our product.

2. We aligned with trends. Text-to-text AI was saturating, and AI was entering the image generation space. We believed one-click text-to-AI memes were perfect, aligning with technological trends and audience demand. This proved crucial.

3. We posted on all our social media channels on launch day to maximize visibility. People who had followed us while we built in public supported us, so our earlier founder-led marketing proved helpful here!

4. We prepared all launch materials in advance and ensured our landing page screenshots precisely conveyed our messaging about easy AI meme generation.

Competing with big dogs

The biggest challenge was sustaining growth when ChatGPT, Anthropic, etc. rolled out native integrations. The traffic and subsequent subscriptions flattened out.

But native integrations lack input and output formats. For example, users need to output quizzes in several formats for learning management systems. These formats are not readily available online, so native apps cannot easily replicate them. So these have been my differentiator.

Launch, iterate, repeat

My advice is to continuously experiment with upcoming models and tools to discover ideas and challenges.

Also, launch lightweight demos and create mini-content on social media platforms like X or YouTube to gauge an idea's vibe and reception.

Then, double down on ideas that work well.

That's exactly what I'm doing. You can follow along on X and LinkedIn. Or check out Questgen.ai, Supermeme.ai, and AiArtist.io.