Arshadul H.

I design NLP systems that work for languages most AI ignores—and ship them to production. I specialize in sentiment analysis, misinformation detection, and low-resource language NLP, particularly Bangla. I've achieved 88% truthfulness accuracy on COVID-19 misinformation detection, built time-series forecasting systems (Kaggle QWK: 0.78), and engineered real estate valuation pipelines handling noisy, unstructured data. I translate between research rigor and business reality. I don't just train models—I handle messy data pipelines, class imbalance, domain-specific terminology, and deployment via Gradio or web interfaces. All work includes open-source code, reproducible pipelines, and documented methodology. Open to: NLP projects (especially multilingual/low-resource), ML system design, prompt engineering, sentiment/sarcasm analysis, and time-series forecasting. Let's discuss your challenge.