Open to opportunities

Kazi Asad

@kaziasad

Data Scientist & NLP Engineer specializing in efficient transformer training, model compression, and reproducible evaluation.

Bangladesh

Message

What I'm looking for

I seek roles applying efficient transformer training and model compression in a collaborative, research-driven environment that values reproducibility and strong evaluation.

I am a Data Scientist and NLP Engineer with hands-on experience building transformer models and classical ML pipelines, focused on efficient training and model compression.

My research and projects include developing LastBERT via knowledge distillation, fake-news detection using SVM with TF–IDF/embeddings, and work on graph vs. hypergraph clustering for protein complex discovery.

I prioritize reproducible experiments and rigorous evaluation (F1, ROC–AUC), and I am proficient in Python, PyTorch, and Hugging Face, with experience deploying lightweight distilled models for resource-constrained settings.

I actively strengthen algorithmic and modeling skills through LeetCode and deep-ml, and I am seeking opportunities where I can apply efficient model training, compression techniques (pruning, quantization, distillation), and strong evaluation practices to real-world NLP and ML problems.

Experience

Work history, roles, and key accomplishments

Teaching Assistant

North South University

Dec 2022 - Oct 2023 (10 months)

Assisted professors with course preparation, conducted tutorials and office hours, graded assignments, and contributed to curriculum development from December 2022 to October 2023.

Tutoring Curriculum Development Grading Teaching Assistance Python

Data Scientist & NLP Engineer

North South University

Developed LastBERT (29M params) via knowledge distillation and led BERT-based pipelines for ADHD severity classification and fake-news detection, achieving competitive performance with substantially reduced model size and inference time.

Python PyTorch Knowledge Distillation NLP TF IDF SVM Model Compression

Education

Degrees, certifications, and relevant coursework

North South University

Master of Science, Computer Science and Engineering

2024 -

Grade: 3.53/4.00

Activities and societies: Research on model compression (pruning, quantization, distillation), graph vs. hypergraph clustering; supervised by Dr. Mohammad Ashrafuzzaman Khan.

Pursuing an M.Sc. in Computer Science and Engineering researching efficient training of large models, Banglish tokenization, and model compression with reproducible experiments and rigorous evaluation.