I'm a researcher at the EPFL NLP Lab in Lausanne, working with Antoine Bosselut on multilingual audio understanding for Apertus 2, Switzerland's fully open multilingual LLM. I'm also a research assistant at the BIIC Lab at National Tsing Hua University, working with Chi-Chun Lee on affective speech foundation models across cultures.
My research is on multilingual and low-resource NLP, mostly around speech: how far synthetic speech can stand in for real data, and the corpora and benchmarks that Persian NLP has been missing. These include ParsVoice, the largest open Persian speech corpus, which I built for my M.Sc. thesis at the University of Tehran with Azadeh Shakery and Heshaam Faili.
Outside research, I love movies and games, and I like karaoke, long walks, exploring random places and visiting museums. You can see what I've been reading on Goodreads. A few of my favorite films and games are hidden around this site as easter eggs; see how many you can find.
I'm looking for PhD positions starting in 2027.
mohammadjavadranjbark@gmail.com/ CV/ Google Scholar/ GitHub/ Hugging Face/ LinkedIn
My thesis model clones a voice it has never heard from a few seconds of audio. Nobody ever recorded this sentence; press the waveform to hear it.
صدای خنده بچهها در تمام کوچه پیچیده بود ▶ press to listen
"The sound of children's laughter echoed throughout the alley."
How far synthetic (TTS) speech can stand in for real speech when pretraining multilingual speech-language models, across six languages.
The largest open Persian speech corpus: 4,000+ hours of audiobooks turned into multi-speaker TTS data by an automated pipeline, with a zero-shot voice-cloning model.
The first Persian audio-language benchmark: 16 tasks and 7,645 samples, from speech recognition to the metre of classical poetry.
20,785 expert-validated questions from 14 years of Iranian medical board exams, in Persian and English, for testing LLMs.
A 17M-sample dataset and a lightweight ParsBERT model for restoring punctuation in Persian text, such as speech-recognition output.
Swiss AI Initiative: adding multilingual audio understanding to Apertus 2, and studying how far synthetic speech can replace real speech in pretraining.
With Chi-Chun Lee, on a Taiwan–Germany project on affective speech foundation models across cultures.
Led research on Persian speech synthesis and medical LLMs; chief TA for ML Methods in NLP and Digital Speech Processing, and TA for five more courses.
A multi-platform RAG customer-service chatbot with 20,000+ active users on the University of Tehran website.
Mentored international students across the NLP, neuroscience and computer vision courses, each summer.
Taught image processing and ML; developed the SWIMBot platform for Diginext; research on facial emotion recognition.
GPA 4.0/4.0. Thesis: Zero-Shot Text-to-Speech for the Persian Language, advised by Azadeh Shakery and Heshaam Faili.
Thesis: Facial Expression Recognition with Robotic Implementation, advised by M. B. Menhaj and H. Taheri.
Audio and Text Understanding for Low-Resource Languages
Audio and Text Understanding for Low-Resource Languages