Bangladeshi engineer wins with PixelFlow

While most people spend their free time scrolling, gaming, or in leisure, Reazul Karim Arif was quietly building skills that would beat over 5,000 global contenders in a high-stakes AI hackathon. Armed with a custom team of AI agents and a single goal to build something real in 24 hours he created PixelFlow. This mood board generator went on to win Lovable's AI Showdown Hackathon, supported by big names like Google, OpenAI, and Anthropic.

Arif's winning entry, PixelFlow, is an AI-powered mood board generator built in just 24 hours. Competing against 5,118 global submissions culled from a pool of 239,826 registered apps-PixelFlow emerged as the final winner, awarding Arif a total of US$20,000 in prize money.

PixelFlow-a glimpse into the future of creative workflows: At its core, PixelFlow transforms abstract ideas into tangible visual inspiration. It was built using React, Vite, TypeScript, Fabric.js, and Huggingface's Transformers-on top of the powerful Claude 4 Large Language Model (LLM) from Anthropic. It enables users to create dynamic, context-aware mood boards using minimal input. The MVP (minimum viable product) was not only technically impressive, but creatively liberating. It's a tool that cuts through creative blocks and streamlines ideation for designers, marketers, or product teams. So it has a huge commercial application in the real world.

However, it wasn't just the product that caught attention. It was the method. Arif designed PixelFlow using a coordinated team of custom AI agentseach responsible for different software engineering roles such as product manager, QA tester, designer, and developer. "In 2025, I can handle different SWE roles with my own custom-built AI agents," Arif said. "They collaborate like a real team, and all I have to do is control them and set different parameters."

Using clever system prompting, employing one-shot or chain-of-thought techniques, he orchestrated these agents to work in parallel. This approach not only increased productivity but became the basis of his entire development process.

From freelance designer to Hackathon hero: Arif's journey so far is anything but conventional. He began as a freelance designer while still in school and later moved to Australia in 2015 to pursue IT studies at LaTrobe University. Now a network engineer for an Australia-based company, Arif's day job doesn't involve building AI agents yet he has been quietly mastering generative AI on the side.

What's more impressive is that the MVP wasn't a result of a year-long dev cycle, it was created in a 24-hour sprint.

Even in a high-stakes global competition with elite SaaS founders like Marc Louvion, John Rush, Tibo Maker, Nick Dobos, and Doctor Yev judging the final round, PixelFlow's depth and execution stood out. First selected by Anthropic for its effective use of Claude 4 and awarded US$10,000, it was later crowned the overall winner, bagging another US$10,000.

A mission to inspire: Despite the global scale of the competition and the sizable prize, Arif's immediate focus is not a big exit or a flashy startup reveal. Instead, he's planning to share his winnings to motivate more Bangladeshis to explore AI.

"I want more people from Bangladesh to get into this new AI-powered coding field," Arif posted. "Dream big. Work hard. This is our future."

He also shared that he plans to polish the PixelFlow prototype and eventually turn it into a full-fledged SaaS platform. The potential is immense: creative tools powered by AI are among the fastest-growing categories globally, from Figma plug-ins to full creative suites like Canva or Adobe Firefly. With the rise of no-code and low-code interfaces, PixelFlow could easily find a niche in this booming ecosystem.

A wake-up call for Bangladesh's tech scene: While India and Southeast Asia have seen increasing representation in the global AI startup ecosystem, Bangladesh is still playing catch-up. Arif's victory is a wake-up call and a proof of concept for the country's vast but often underutilised tech talent. Bangladesh has no shortage of capable engineers, designers, and coders. What's often missing is exposure, mentorship, and the push to move beyond routine IT jobs or freelance marketplaces. Platforms like Lovable, OpenAI's hackathons, and global competitions offer exactly the kind of sandbox where Bangladeshi innovators can test ideas, build with frontier tools, and get real-time feedback from international judges.

Arif's story is a clear reminder that with a commitment to learning, smart experimentation, and clever prompting, the global tech field is wide open. The future of work is being co-written by AI agents and humans alike. Those who understand how to leverage this collaboration will lead the next wave of innovation. "It's not rocket science," Arif says. "Just clever system prompting."

If more Bangladeshis take that message seriously, we won't just be outsourcing talent, we'll be exporting innovation.

outreachfe@gmail.com