Speech emotion recognition
Can a model hear emotion in a stranger's voice? Less than it seems: 65.1% fell to 58.0% on unseen speakers.
A story told in sound
Every sound begins in silence.
01
I grew up in Yazd, a city of windcatchers in the middle of the desert.
In sixth grade, a great teacher opened a door to NODET, Iran's schools for gifted students. From my class, only I walked through.
Six years there. The first three were math olympiads, contests and algorithms, mostly to prove a point to myself.
02
In tenth grade, I found music.
Classical piano first, third place in Yazd. Then my own tracks in FL Studio, and an album in eleventh grade.
Music didn't become my path. It pointed to one, the way these strings do: engineering, where sound becomes signals.
This is Lost: a teenager's experiment, kept just as it was.
03
On paper, I was headed for medicine.
In practice, I kept building things. In ninth grade, with two friends beside me, I started Bartarinha, a team math competition between our school and our sister school.
It ate almost the whole budget. First business lesson: passion lights the fire, money keeps it burning.
In the year everyone studied for the national entrance exam, I stepped off the road. A year of talking to people and digging around, and I landed on computer engineering.
A few months of studying later, I was in. University of Isfahan.
04
I walked in with almost no code behind me.
So I started where everyone starts: hello, world. Then I took two roads at once.
Run your hand through the curtain.Touch the curtain.
Two roads
Product on one side, AI on the other. By the end, one road. Tap a course.
Melodies as language
Persian melodies read as language: n-gram models finish the phrase, quarter-tones and all.
Local research assistant
Answers drawn from research papers, each citing its page, running on one laptop.
- PDF pages
- chunks
- MiniLM
- FAISS
- Qwen2.5
- answer + page
Persian search engine
Search for Persian text, ranked with BM25.
Hazm · inverted index · BM2506
Say hello.
Talk, hum, sing. The sphere answers in key.
Your voice stays in your browser. Nothing is recorded or sent.
07
Next, I want to go deeper.
A research MSc or a direct PhD in speech and audio machine learning. Until then, I'm open to applied AI roles, especially in NLP, speech and signal processing. If that sounds like you, my inbox is open.
moeinbagheri.ai@gmail.com
Write to me
