Skip to content
Abdullah Al Rafi

About

From production ML at bKash to an MSc in NLP.

I’m an ML engineer who enjoys the whole path from a model to a product: training it, serving it, and building the APIs and interfaces around it.

At bKash, Bangladesh’s largest mobile-financial service, I built a first-of-its-kind OCR for Bangladeshi birth certificates that reads Bangla and English, served it on NVIDIA Triton at 100k+ requests a week, and later extended it to new document types, which lifted weekly successful onboarding up to 3×. Before that I fine-tuned Bengali speech recognition with NVIDIA NeMo at Aequisym AI.

Today I’m studying for an MSc in Natural Language Processing at Universität Trier. I’m looking for Werkstudent and thesis roles where NLP meets production, especially document AI and retrieval. Outside of work I’m curious about open source, security and finance.

Read the case studies or get in touch.

Education

  1. MSc Natural Language Processing

    Universität Trier, Trier, Germany

    2025 – present

  2. BSc Computer Science & Engineering

    North South University, Dhaka, Bangladesh

    2017 – 2022

Publications

  1. Use of Social Media in Flood Assessment in Bangladesh

    F. H. Shubho, E. Sarker, A. Al Rafi, P. Rahman, M. T. M. Habib, T. Ihsan, R. M. Rahman

    2022 IEEE 11th International Conference on Intelligent Systems (IS), 2022

    DOI 10.1109/IS57118.2022.10019640

Languages

Bangla
Native
English
Professional working proficiency

Interests

NLP & LLMs · Document AI · Deep learning · Open source · Finance