Started in a cramped Kaohsiung apartment back in 2019, FlashStream emerged from frustration. Not with technology—with how people taught it. We wanted to create something that felt more like solving puzzles with friends than attending lectures.
Five years sounds short. But when you're building education from scratch, every month teaches you something unexpected about how people actually learn.
Three developers, one shared desk, and an idea that probably seemed naive at the time. We noticed students could build neural networks but couldn't explain why they worked. So we started hosting weekend workshops in coffee shops. Turns out, people wanted that.
Got our own space on Zhongzheng 3rd Rd. Launched our first structured program. Learned that twelve-week courses don't work for working professionals. Cut everything down to six weeks with flexible scheduling. Enrollment doubled.
Added online cohorts because people kept asking. Worried it would feel impersonal. Instead, students from Taipei, Singapore, and Tokyo started collaborating on projects. That cross-pollination of perspectives became one of our strongest features.
Now we're focused on specialized tracks—computer vision, NLP, reinforcement learning. Our autumn 2025 cohort will test a mentorship model where experienced learners guide newer ones. We think peer learning beats traditional instruction for practical skills.
These aren't values we brainstormed in a meeting room. They're patterns that emerged from watching hundreds of students figure out deep learning.
Theory matters, but nothing teaches like debugging your own broken model at 2 AM. We prioritize hands-on projects over perfect explanations.
The best insights come from group debugging sessions. We create environments where asking "dumb questions" is encouraged, not just tolerated.
We don't simplify frameworks or hide complexity. You'll work with the same messy, powerful tools professionals use daily.
Your first model will be terrible. That's expected. We measure growth by how quickly you can iterate, not how polished your initial attempt looks.
Small team, diverse backgrounds. What connects us is we all learned deep learning the hard way and want to make it easier for others.
Lead Instructor
Former researcher who realized she enjoyed teaching more than publishing. Specializes in making complex architectures understandable without dumbing them down.
We don't claim to have perfected education. Each cohort teaches us something new about how people learn differently. What we can promise is that we'll keep experimenting, listening to feedback, and adjusting our approach. If that sounds like the kind of environment where you'd thrive, we should talk.
Get In Touch