Forem: Meheret Egzerab

I Built an AI Tool That Creates App Store Screenshots in Seconds. Here's How Google Gemini Made It Possible

Meheret Egzerab — Tue, 03 Mar 2026 23:06:23 +0000

This is a submission for the Built with Google Gemini: Writing Challenge

What I Built with Google Gemini

As a solo indie developer shipping iOS apps under my studio, Studio Rehoboth, I kept running into the same bottleneck every time I was ready to launch: App Store screenshots.

Apple requires very specific image dimensions (1284 x 2778 px for iPhone, 2048 x 2732 px for iPad), and making them look professional with device mockups, catchy headlines, and matching color palettes used to mean hours in Figma. For every single app. Every single update.

So I built AppStoreGenius: a personal AI-powered tool that turns my raw app screenshots into polished, Apple-compliant App Store assets in seconds. I built it directly in Google AI Studio using the Gemini API, and it's become an essential part of my development workflow.

Here's what it does:

AI-Powered Design Generation: I upload a raw screenshot of my app, and Gemini analyzes the image to automatically generate a marketing headline, descriptive subtitle, and a color palette (background + text colors) that matches the app's branding.

Smart Status Bar Masking: The AI identifies the exact color of the app's top navigation bar and applies a matching mask to hide the real status bar icons (battery, carrier, time), giving that clean, professional look.

Automatic Device Mockups: The screenshot gets placed inside a realistic device frame, complete with bezels and Dynamic Island for iPhone.

Full Customization + One-Click Export: I can tweak headlines, adjust backgrounds, change device types, and fine-tune positioning. When I'm happy with it, I export everything as high-resolution PNGs at Apple's exact required dimensions, packaged in a ZIP file.

I've used AppStoreGenius for two of my apps so far:

Lent Journey: A Catholic devotional app for Lent that hit #1 in the App Store for "lent." The professional-looking screenshots absolutely helped with conversions.
Advent Habit Calendar: A faith-based habit tracking app where I used the same workflow to generate App Store assets quickly.

Demo

The source code is on GitHub: github.com/mercyg/App-store-genius

It's built with TypeScript and React, generated from the google-gemini/aistudio-repository-template.

The tech stack:

Frontend: TypeScript + React (Vite)
AI: Google Gemini API for image analysis, headline generation, and color extraction
Export: Client-side canvas rendering for pixel-perfect output at Apple's required dimensions

The tool isn't publicly deployed yet; right now, it's part of my personal development toolkit. But the code is open source if you want to see how it works under the hood.

What I Learned

The biggest lesson: AI tools are most powerful when they solve YOUR specific pain point. I didn't build AppStoreGenius as a product idea. I built it because I was tired of spending hours in Figma every time I shipped an app. The fact that it works so well is a testament to how capable the Gemini API is for practical, workflow-level automation.

Gemini's vision capabilities are genuinely impressive. The fact that it can analyze a screenshot, identify the dominant colors, understand the context of the app, and generate relevant marketing copy all from a single image upload still surprises me. The color extraction for status bar masking works remarkably well.

AI Studio made prototyping fast. Being able to iterate on prompts and build the tool directly in the AI Studio interface meant I went from idea to working app quickly. And now every time I need App Store screenshots, I just upload my raw screenshots, let Gemini do its thing, hit export, and get a ZIP file with everything at the exact dimensions Apple requires, ready to upload straight to App Store Connect.

Professional presentation matters more than you think. When I launched Lent Journey, I genuinely believe the polished App Store screenshots contributed to the strong conversion rate. People judge apps by their screenshots before they ever download, and having AI-generated professional assets lets me compete with apps that have dedicated design teams.

Google Gemini Feedback

What worked well:

The image analysis capabilities are the star here. Uploading a screenshot and getting back a coherent headline, subtitle, and color palette that actually matches the app's vibe — that's incredibly useful. For Lent Journey, it nailed the spiritual, contemplative tone. The color detection was accurate enough that the status bar mask blended seamlessly.

The AI Studio prototyping experience was also smooth. Being able to iterate on prompts and see results immediately, then export to a proper codebase, meant I went from idea to working tool faster than I expected.

Where I ran into friction:

Getting the exact output format right took multiple iterations of prompting. Apple's requirements are very specific — exact pixel dimensions, exact aspect ratios — and the initial prompts didn't always respect those constraints. I had to be very explicit about the output specifications, which required some back-and-forth.

The Gemini API sometimes generated headlines that were a bit generic or didn't quite capture the unique value proposition of the app. It's great as a starting point, but I almost always tweaked the generated copy. This is fine for my workflow since the tool supports manual editing, but the AI-generated marketing copy does need a human eye.

Overall, Gemini is at its best when you give it a focused, well-defined task, analyze this image, extract these colors, generate a headline in this style — rather than asking it to handle the entire creative process end-to-end. That's exactly how I architected AppStoreGenius, and it works beautifully.

I'm considering publishing AppStoreGenius as a tool that other indie devs can use. If that's something you'd find useful, let me know in the comments! And if you've built something cool with Gemini, I'd love to hear about it.

What If You Learned Data Structures by Building Real Projects?

Meheret Egzerab — Wed, 25 Feb 2026 16:58:16 +0000

Most DSA resources follow the same pattern, here's a data structure, here's how it works, now solve this isolated problem. Textbooks, LeetCode, YouTube courses, the format changes but the approach is pretty much the same.

I wanted to try something different. What if you learned data structures by building a real project and the DSA concepts came up naturally because you actually needed them?

So I built BuildCode, a free platform where you learn data structures by building real projects. It's an experiment, and I want to see if this approach clicks for people.

The idea is simple

Instead of "Given an array of integers, find two numbers that add up to a target..." you get:

"Build a Task Manager. Along the way, you'll discover why hash maps exist and when they're the right tool."

You start with a working project. You write real code. You hit real performance problems. And then the data structure clicks not because you memorized it, but because you needed it.

What's live right now

The first lesson is a Task Manager project that teaches hash maps through 10 hands-on steps. It's a mix of:

Running pre-written code to see concepts in action
Writing key portions yourself (not just clicking "next")
A progressive hint system when you get stuck so you're challenged but never stranded

It uses modern JavaScript (Map, Set) the stuff you'd actually use at work, not a from-scratch implementation you'd never ship.

Why I think this matters more now, not less

Hot take: AI making code generation easier makes understanding fundamentals more important. If you don't understand why a hash map is the right choice, you can't evaluate whether Copilot gave you a good solution or a terrible one. You just copy-paste and hope.

BuildCode is about building the intuition that AI can't give you.

Try it and tell me what it's like

The Task Manager lesson is free, no signup wall, just go try it:
👉 https://www.buildcode.codes/
I'd genuinely love feedback. Things I'm especially curious about:

Did the progression make sense? Or did any step feel like a jump?
Was there enough hands-on coding? Or did it feel too guided?
Would you want more projects like this? (Graphs → social feed, tries → autocomplete, etc.)

Drop a comment here, open an issue, or DM me, whatever's easiest. I'm building this in public and real feedback is everything.

If you're an educator using project-based approaches to teach DSA, I'd especially love to hear from you.

Solo Command:- A mindful standup tool built for solo developers

Meheret Egzerab — Mon, 16 Feb 2026 04:52:24 +0000

This is a submission for the GitHub Copilot CLI Challenge

What I Built

Solo Command is a web-based daily standup and project tracking tool built specifically for solo developers and indie founders.

Here's the problem: when you're a solo founder, there's no morning standup. No one asks what you did yesterday, what you're working on today, or what's blocking you. Days blur together. You context-switch between projects and lose track of progress. Then, when you need to write a weekly update, pitch an investor, or remember what you did last Tuesday, you're scrambling.

I built Solo Command because I'm living this problem right now. I'm a solo founder juggling multiple app launches, hackathons, and marketing — all at once. I needed a calm, structured space to capture my daily progress across everything.

Key features:

📝 Daily Standups - Log what you accomplished, what's next, and what's blocking you
💡 Idea Capture — Quickly save ideas without breaking your flow
🎯 Multi-Project Tracking — Tag entries by project, filter, and see momentum across all your work
📊 Timeline & Board Views — Toggle between a chronological timeline and a visual board layout
🌅 Mindful Design — A warm, intentional interface that feels like a sanctuary, not another noisy productivity tool

Demo

🔗 Live App: solo-command

🔗 GitHub Repo: github-repo

My Experience with GitHub Copilot CLI

I built Solo Command almost entirely through Copilot CLI over a weekend. Here's how it shaped the process:

Architecture & scaffolding — I started with just a concept. Copilot CLI helped me pick the tech stack, set up the project structure, and generate the initial codebase in one conversation.
Feature development — The standup system, project tagging, timeline/board views, and idea capture were all built through iterative prompts. Describe → review → refine. Repeat.
Debugging — When things broke, I stayed in the terminal. Copilot CLI gave me targeted fixes without the usual context-switching to docs and Stack Overflow.

The biggest surprise was how natural it felt. Copilot CLI handled the full lifecycle — not just code generation, but architectural decisions and problem-solving. It met me where I already work: the terminal.

Built in a weekend by a solo founder, for solo founders. Because even when there's no team, you still deserve a standup. ⚡

— Hackathon — What I learned and why you should attend!

Meheret Egzerab — Wed, 24 Feb 2021 05:48:03 +0000

This post was originally published in Medium in 2017.

Over the past few days, I was going through my journal and found this interesting note about my first Hackathon experience. I wrote this note for myself in hopes of sharing my experience with someone like you one day:-

I had the chance to attend my first Hackathon one year ago. It was a great experience. I got to meet with different people and hear about their coding journeys. I also met with people who don’t code at all but have great ideas about the kind of apps that they’d like to create someday. These people in particular came to the hackathon to collaborate and work with other people who can program so that they could publish an app or project. The hackathon was held in DC by Women Who Code, an organization that empowers women in the tech world.

I arrived in the morning and after a few minutes-, the program officially started and I started to look around the project that was listed off. At the time, there were so many great ideas it was hard to decide which project to get involved with. I finally decided to join a team with an emergency management idea. Long story short, the project goal was to improve the management system for Virginia state’s emergency department. We started off by introducing ourselves and continued to brainstorm on how we could improve the department. With so many ideas and because of time constraints we ended up only creating a mock-up for the website for the department in the end.

The main reason I’m sharing my experience is not to talk about the step-by-step process of my participation in the hackathon, but rather the insight I gained from attending. You see, I have been teaching myself to code for the past months and this was the first time I was sharing what I know with others. Attending the hackathon, gave me the realization that all the time I spend staring at my laptop is actually a big deal. When I started sharing what I know with other people, I realized that what I have been doing has value, and that was such a good feeling.

Most of the time when I was learning by myself even after finishing courses online I would highly underestimate what I had accomplished. I would think that what I know now was still not enough and focus on the fact that there was a lot more still to learn. Honestly, attending the hackathon helped me to realize that I really have gained a lot of knowledge and know a lot already, more than I thought I knew.

“Trust yourself, you know more than you think you do.” Benjamin Spock

Though learning to code, I’ve found that it’s good to admit what I know and to be proud of it. This has given me the motivation to keep on learning and doing more. On top of that, the hackathon showed me different kinds of learning processes and tools for solving problems that other people use. I also learned how great it is to work in teams, which has additionally has made me think outside of the box.

When I truly think about the knowledge and the experience that I got from the hackathon, -I thought it would be a great motivation for anyone else who wants to attend one, and for this reason, again, I’m sharing my experience. It might be scary when you think about attending a Hackathon for the first time but once you get there, I know you won’t regret it. So go ahead and search for one in your city or country and attend!!!
Thank you for taking the time to read this. If you liked it, click the heart button so other people will see it here on DEV.

Happy coding!

Nevertheless, Meheret Girma Coded

Meheret Egzerab — Wed, 08 Mar 2017 01:01:04 +0000

I began coding because...

I was introduced to coding when I took my first computer science class. When I started I was not encouraged or happy by coding. But after learning for a while and coding more I started to love it more.

I'm currently hacking on...

I have started working on a side project to build a web application that lets users review there coffee and hope to make it a mobile app!

I'm excited about...

New ideas and helping people!

My advice for other women who code is...

Keeping pushing and doing what you love! Even if the road is rough just keeping doing what you love. We can do big thing if we encourage eachother!