- 🛢️ I specialize in SQL & Database Engineering — temporal systems, schema design, and query optimization
- 🤖 I build ML systems — from NLP to production-grade MLOps pipelines
- ☕ I work with Java Spring Boot, Python FastAPI, and C/C++ for backend & systems programming
- 🔬 My B.Tech thesis introduced HNVec — a custom Hindi NLP embedding achieving 88.61% accuracy
- 🤝 I enjoy collaborating on complex, end-to-end systems
This project is a Java Spring Boot and MySQL-backed REST API implementing a Temporal Database System to track historical data states. It uses a JAXB-based schema compiler to automatically generate history tables from XML metadata and supports complex queries like temporal joins (executeWhen) and time-slices (executeTimeSlice) to audit and retrieve past database records.
MySQL XML Java Spring Boot
🔗 https://github.com/devdeepsarkar/TemporalDBSystem
This system analyzes chat conversations (like WhatsApp exports) using two fine-tuned BERT models — one to detect if a message is cyberbullying, and another to classify it's subtype (age, gender, religion, ethnicity, or other). It goes beyond simple text classification by incorporating relationship analysis (friends vs. strangers), a sliding 4-message context window, victim rescue logic, and regex guardrails to accurately identify aggressors, targets, and severity level in real-world conversations.
NLP Python PyTorch HuggingFace scikit-learn Flask
🔗 https://github.com/devdeepsarkar/ContextAwareCyberBullyingDetectionSystem
Designed HNVec, a domain-specific Hindi word embedding model, and integrated it with hybrid CNN–LR–SVM architectures for large-scale Hindi news classification, boosting accuracy from 70% to 90% compared to TF-IDF and FastText approaches.
Python TensorFlow Scikit-Learn NLP
🔗 https://github.com/devdeepsarkar/Project-HNVec
This project is an end-to-end, production-grade cybersecurity MLOps pipeline that uses FastAPI and Streamlit to serve and visualize Random Forest and Autoencoder models for detecting network attacks and zero-day threats. It features a self-healing architecture automated via Jenkins and Ansible, orchestrated on Kubernetes, monitored by the ELK stack, and equipped with a drift detection service that automatically triggers model retraining and hot-reloads during traffic distribution shifts.
Docker Kubernetes Ansible Jenkins ELK Stack FastAPI Streamlit Python Microservice MLOps
🔗 https://github.com/devdeepsarkar/Drift-Aware-MLOps-Platform-for-Cybersecurity-Anomaly-Detection
Client-server banking system handling accounts, loans, transactions, employees, and management operations.
C File I/O Linux(Ubuntu)
🔗 https://github.com/devdeepsarkar/BankingManagementSystem_SS
ERP module for academic salary management with backend services and relational database integration.
Spring Boot React TypeScript MySQL
🔗 https://github.com/devdeepsarkar/Academic-erp-salary-disbursement