Portrait of Amirreza Sattarzadeh Khanehbargh

Amirreza Sattarzadeh Khanehbargh

M.Sc. student and Graduate Research Assistant · Text Analytics and Bioinformatics Lab (TabiLab), Boğaziçi University, Istanbul

I work on natural language processing and machine learning, centred on language models: how their input units (tokenization) are chosen, how they are adapted and fine-tuned, and what their representations encode. I am especially interested in bringing NLP and ML into other fields, such as biology, social science, finance and low-resource, less-studied languages. My current research, in the ERC-funded LifeLU project with Prof. Arzucan Özgür, treats proteins as a language and studies tokenization and adaptation in protein language models. I have also recently joined the PARI project, working on hate speech detection in Turkish and Arabic. PARI is developed by the Hrant Dink Foundation with Boğaziçi and Sabancı universities and funded by the European Union and the Friedrich Naumann Foundation.

Before Boğaziçi I worked on financial NLP at Iran University of Science and Technology, where I was also a teaching assistant for seven course sections, including Head TA for Data Science and Scrum Master in two software engineering courses. In industry I built speech and text data pipelines for a Persian speech recognition product.

Looking for PhD positions starting in autumn 2027
CV (PDF) Email Google Scholar GitHub LinkedIn

Happy to discuss PhD opportunities: amirreza.sattarzadeh@bogazici.edu.tr

Research interests

Natural Language Processing (NLP) and Machine Learning (ML), centred on language models (LMs): tokenization and input representations, efficient adaptation and fine-tuning, and interpretability of what models encode.

Bringing NLP and ML to other fields. I want to combine NLP and ML with other domains and their questions: biology (protein language models and BioNLP), the social sciences (hate speech in multilingual text), finance (market-derived sentiment), and low-resource, less-studied languages and their linguistic structure, for example through Universal Dependencies (UD) treebanks (I have worked with Turkish UD data).

Core
Natural language processing Machine learning Deep learning Data science Language models Tokenization Efficient adaptation & fine-tuning Interpretability
Applied to
BioNLP Protein language models Social NLP & hate speech Financial NLP Multilingual & low-resource NLP Universal Dependencies Privacy of synthetic data

News

Show older news

Publications and posters

  1. Benchmarking Recent Single-Cell Synthetic Data Generation Methods on the OneK1K Dataset Buse Giledereli*, Başar Temiz*, Amirreza Sattarzadeh Khanehbargh* ELSA Health Privacy Challenge, CAMDA 2026 at ISMB 2026, Washington, DC (selected talk and poster) · Preprint, 2026 · *equal contribution
    arXiv Challenge

    We benchmark single-cell RNA-seq generators, including an LLM-based one (Cell2Sentence), for biological utility against membership inference privacy, and show that how non-highly variable genes are reconstructed is the main privacy lever.

  2. Tokenization-Aware Protein Language Modeling with Evolution-Guided Units and Biological Knowledge Fusion Amirreza Sattarzadeh Khanehbargh, Burak Suyunu, Özdeniz Dolu, Arzucan Özgür 25th European Conference on Computational Biology (ECCB 2026), Geneva · Poster C-P.51, presented in person
    Poster page
  3. From Protein Sequences to Biological Language Units: Frequency- and Evolution-Aware Tokenization for Protein Language Models Amirreza Sattarzadeh Khanehbargh, Arzucan Özgür Lectures on Computational Linguistics (LCS 2026), Bolzano, Italy · Accepted poster
  4. Market-Derived Financial Sentiment Analysis: Context-Aware Language Models for Crypto Forecasting Hamid Moradi-Kamali, Mohammad-Hossein Rajabi-Ghozlou, Mahdi Ghazavi, Ali Soltani, Amirreza Sattarzadeh, Reza Entezari-Maleki Preprint, 2025 · 12 citations on Google Scholar, including its earlier version (Oct 2026)
    arXiv

    Labels social media posts by the market move that follows them instead of human sentiment labels and fine-tunes domain language models on these labels: up to +11% trend prediction accuracy, and 89.6% with market and temporal context in the prompt.

Research experience

Graduate Research Assistant, LifeLU (ERC Consolidator Grant)

Oct 2024 – present

TabiLab, Boğaziçi University · PI: Prof. Arzucan Özgür · Understanding the Language of Life: Identifying and Characterizing the Language Units in Protein Sequences

  • Tokenization and fine-tuning of protein language models, including evolution-aware tokens and added biological knowledge.
  • Next: interpreting the learned units, other knowledge fusion methods and their interaction with tokenization, and dynamic and task-specific tokenization.
  • Regular presenter in the team's weekly reading club.

Developer and Researcher, PARI project

Oct 2026 – present

Developed by the Hrant Dink Foundation in collaboration with Boğaziçi University and Sabancı University, Istanbul · funded by the European Union and the Friedrich Naumann Foundation

  • Recently joined as the project's student researcher, working with its faculty, postdoctoral and PhD researchers on an open-source NLP tool for detecting hate speech and discriminatory content in Turkish and Arabic.

Team member, ELSA Health Privacy Challenge, CAMDA 2026 at ISMB 2026

2026

Organized by EMBL, CISPA Helmholtz Center and the University of Helsinki within the EU ELSA project

  • Equal-contribution first author with Buse Giledereli and Başar Temiz. Benchmarked single-cell synthetic data generators, including an LLM-based one, for biological utility versus privacy.
  • Selected for a talk at CAMDA 2026 and presented as a talk and a poster at ISMB 2026, Washington, DC (July 2026).

Research Assistant

Jul 2023 – Sep 2024

Iran University of Science and Technology (IUST), Tehran · Supervisor: Dr. Reza Entezari-Maleki

  • Financial market trend analysis using FinBERT and sentiment analysis in online text data (Feb – Sep 2024), extended into the arXiv preprint.
  • Deep learning and neural networks for algorithmic trading (Jul 2023 – Feb 2024).

Industry experience

Python Developer (Backend and Data), Roshan-AI

Jun 2022 – Nov 2023

Tehran · deep learning company · joined through a competitive backend internship, then part-time developer

  • Harf (Persian speech recognition): built the data pipeline for crawling, cleaning and preprocessing speech–text data, growing the training data in size and quality, which improved the speech models; Django APIs for model deployment.
  • Replai (NLP comment and reply classification): Django backend.
  • Recommendation from my team lead (LinkedIn, March 2024):

Junior Software Engineer (freelance)

Sep 2019 – Feb 2022

Bonab, Iran

  • Telegram bot that crawls Iranian stock market data and screens growth stocks.
  • Django backend for a charity website, in a team of five.
Earlier roles2019 – 2024
  • IraCode, Tehran: International Cooperation Adviser for Türkiye and Azerbaijan, and PHP intern (Mar – Jul 2020).
  • MovieBaz: Turkish-to-Persian translator and subtitler (Mar – Aug 2021).
  • AzmaFanParvaran: WordPress developer (Jan – Jul 2019).
  • Independent trader and investor in equity and crypto markets (2019 – 2024), the motivation behind my financial NLP work.

Teaching and mentoring

Teaching Assistant, Iran University of Science and Technology (IUST)

Sep 2023 – Jul 2024

Seven course sections across two semesters

  • Head TA, Data Science (Dr. Hassan Naderi), Fall 2023.
  • Software Engineering Mentor and Scrum Master, Systems Design and Analysis (Dr. Mehrdad Ashtiani), Fall 2023.
  • Scrum Master, Software Engineering II (Dr. Behrouz Minaei-Bidgoli, Dr. Mehrdad Ashtiani), Fall 2023.
  • TA, Fundamentals of Computer Programming (Dr. Marzieh Maleki-Majd), Fall 2023.
  • TA, Computer Workshop (Dr. Marzieh Maleki-Majd), Fall 2023.
  • TA, Advanced Programming, two sections (Dr. Marzieh Maleki-Majd, Dr. Marzieh Davoudabadi), Spring 2024.

Python Instructor (independent, online)

Jan 2023 – Jun 2024
  • Taught introductory Python programming to beginners.

Volunteer Instructor, Pooyandegan-e-Daneshpazhooh-e-Mehr (BonabNGO)

Summer 2023
  • Designed and ran a free online Python and computer engineering fundamentals course to reduce educational inequality in the region.

Education

M.Sc. in Computer Engineering, Boğaziçi University

Oct 2024 – expected Jul 2027

Istanbul, Türkiye · English-taught · Supervisor: Prof. Arzucan Özgür

  • Thesis (in progress): From Protein Sequences to Biological Language Units: Frequency- and Evolution-Aware Tokenization for Protein Language Models.

B.Sc. in Computer Engineering, Iran University of Science and Technology (IUST)

Sep 2018 – Sep 2024

Tehran, Iran

  • Final project: Financial Market Trend Analysis using Sentiment Polarity in Online Text Data (Dr. Reza Entezari-Maleki), extended into the arXiv preprint.

High school diploma, Mathematics and Physics

2014 – 2018

SAMPAD (NODET: National Organization for Development of Exceptional Talents), selective school for gifted students, Bonab, Iran

  • Ranked 1st among the school's Mathematics and Physics students.

Honors and tests

Community and service

Beyond research

Music

Bağlama player in the Zakhme Bonab Youth Music Group (2020 – 2022).

Speed skating

1st place (2017) and 2nd place (2019) in Bonab County competitions.

Taekwondo

Silver medal, Bonab County selection (2016).

Languages

Persian, Azerbaijani and Turkish (native or bilingual); English (full professional; English-taught M.Sc.); Arabic (elementary). Turkish-to-Persian translator and subtitler for MovieBaz (2021).