About
Tanbir Hossain Ramim
Software engineer in Regensburg, Germany, previously building from Dhaka. Specializing in full stack development, cloud infrastructure, and AI automation systems.

The long version
I started with C, out of an ICT textbook, in a college classroom in Dhaka. It was supposed to be one chapter of a syllabus. I found it more interesting than the rest of that syllabus put together, and that turned out to be the whole decision.
What followed was competitive programming, which I recommend to anyone who wants to learn quickly and does not mind being wrong in public. Over thirty contests and hackathons like ICPC, IUPC, university rounds, later Hackaburg and Hack | Bay here in Germany. A wrong answer verdict does not explain itself. You learn to find your own mistakes, and that is most of the job.
In 2023 I paused everything to move countries. Between May 2023 and June 2024 I was not building anything: paperwork, exams, and waiting on a visa to study in Germany. It is the one real gap in my CV, and it seemed better to account for it here.
Since then: a computer science degree at OTH Regensburg, a working student job rebuilding a hotel group's web estate and the operations software behind it, and a student assistant role building a voice negotiation simulator for Sales Management students.
The direction I am moving in is AI engineering, and I did not want to claim that on a CV without something behind it. So I built EnAi. It watches a training run in flight, retires layers once their gradients collapse, and cuts the backward compute mid-run. It removes 6.91% of training compute while touching 0.35% of the parameters.
The part I care most about is the part most efficiency projects skip: the page listing what the measurement does not establish. The energy saving is not proven, because the meter was estimating, and accuracy parity is not proven either. Both belong next to the result rather than somewhere further down the page.
Away from all of that I climb, play chess, and play guitar. I also work on Phoveus Lab with a few other people, building AI products that we give away free and open source.
How I work
What I bring to a team
Measure it, or do not claim it
The EnAi study has a section listing what the numbers do not establish, because an efficiency claim is only worth what its methodology survives. Any result I quote, I can show you how it was taken.
Latency is a feature, not a metric
The negotiation simulator lives or dies on response time. A two second pause turns a conversation into a form submission. I would rather ship something slightly less clever that answers immediately.
The boring half is usually the valuable half
Guests see the booking engine. Staff see the scheduling tool nobody demos. The second one is what actually gave people their week back.
Free where it matters
The people who most need software that removes drudgery are usually the least able to pay for it. That is why the Phoveus Lab work is open source and stays that way.