From Research to Production: Deploying an AI-Powered Bangla News Classifier
A complete journey from research paper to production deployment using Django, Docker, CI/CD pipeline and DevOps practices.
Mahedi Hasan EmonApril 4, 20262 min read
🚀 From Research to Production: Deploying an AI-Powered Bangla News Classifier 🇧🇩 Six months ago, I published a research paper at IDAA 2025. Today, that same research is live on the internet. Fully deployed, secured, and accessible to anyone. Here’s the full journey 👇 --- 📄 The Research I published: “Enhancing Bangla Document Classification Using a Hybrid Ensemble of Bangla-BERT and Bi-LSTM Models” at the *International Conference on Intelligent Data Analysis and Applications (IDAA 2025)*. The approach: • Bangla-BERT → contextual understanding • Bi-LSTM → sequential learning • Hybrid ensemble → smarter predictions --- ⚙️ The Build After publication, I asked myself: *Why let this sit in a PDF?* So I rebuilt everything from scratch. Using Django, I turned the research into a real application where users can: • Input Bangla text • Get instant classification • See confidence scores Now it’s not just research, it’s a usable product. --- 🔁 The Automation (CI/CD) I didn’t want manual deployment. So I implemented a pipeline using GitHub Actions: ✅ Auto-build Docker image on every push ✅ Push to Docker Hub ✅ Fully automated, reproducible deployments --- ☁️ The Deployment (Production Infrastructure) For production infrastructure, I used: 🖥 DigitalOcean VPS - Ubuntu 24.04 (SFO3) 🐳 Docker - containerized Django + ML stack 📦 Portainer - visual Docker management 🔀 Nginx Proxy Manager - reverse proxy with SSL 🌐 Cloudflare - DNS, DDoS protection, subdomain routing --- 🌐 Live System 🔗 Live Project: https://newsclassifier.mahedihasanemon.site 🌐 Portfolio: https://mahedihasanemon.site --- 🧠 What I Really Learned This project completely changed how I see development: - ML/DL alone isn’t enough - Web development makes it usable - DevOps makes it scalable and real I also faced real-world problems along the way: • CSRF errors • Container issues • Port conflicts • Reverse proxy configuration • Domain & SSL setup …and honestly, that’s where the real learning happened. --- 🛠 Tech Stack Python • Django • Bangla-BERT • Bi-LSTM TensorFlow • PyTorch • Docker • GitHub Actions Nginx • Cloudflare • DigitalOcean --- If you're a researcher or ML engineer and your work is still sitting in a notebook or paper… 👉 Try deploying it 👉 Let real users interact with it That’s where real engineering begins. --- Happy to discuss the ML architecture, backend, or deployment pipeline, feel free to reach out 👇 #OpenToWork #DevOps #MLOps #MachineLearning #DeepLearning #NLP #BanglaNLP #WebDevelopment #Django #Docker #GitHubActions #CICD #CloudComputing #SoftwareEngineering #AIEngineering #ResearchToProduction