I'm Md Mazidul Islam, a backend engineer interested in building reliable software systems and practical AI-powered applications. I have 1.5+ years of backend development experience and a growing focus on ML engineering, LLM applications, and intelligent automation.
- βοΈ I build backend services, APIs, database-backed applications, and asynchronous workflows.
- π§ I explore RAG, LLM evaluation, AI agents, model serving, and ML system design.
- π‘οΈ I value measurable results, testing, data quality, security, and clear engineering trade-offs.
- π B.Tech in Computer Science and Engineering β Lovely Professional University (2020β2024).
My engineering approach: make it work, measure it, test its limits, and make it reliable.
|
Backend Systems APIs Β· Databases Β· Queues |
Agentic AI LLMs Β· RAG Β· Automation |
ML Engineering Evaluation Β· Serving Β· Drift |
Distributed Workflows Redis Β· Workers Β· WebSockets |
π§ SupportIQEvaluated, guardrailed RAG service A policy-grounded support assistant with hybrid retrieval, citation checks, prompt-injection filtering, PII redaction, caching, and an offline evaluation harness. Python Β· FastAPI Β· BM25 Β· MiniLM Β· Hugging Face Β· pytest Explore repository β |
π RetainIQStudent churn prediction & monitoring A time-aware ML pipeline with leakage tests, risk ranking, SHAP explanations, a FastAPI service, and PSI-based drift monitoring. Python Β· SQL Β· scikit-learn Β· XGBoost Β· SHAP Β· MLflow Β· FastAPI Explore repository β |
π€ Agentic HRAI-powered recruitment platform An event-driven recruitment ecosystem covering CV/JD screening, candidate assessments, interview scheduling, AI voice/video interviews, and candidate data lifecycle workflows. NestJS Β· FastAPI Β· Next.js Β· PostgreSQL Β· Redis Β· Groq Β· Docker Explore repository β |
Queue-driven video processing backend A media pipeline for audio extraction, AI transcription, topic segmentation, automatic clipping, background processing, and temporary-file cleanup. Node.js Β· Express Β· MySQL Β· Redis Β· Bull Β· FFmpeg Β· Whisper Β· Claude Explore repository β |
Note: SupportIQ and RetainIQ use synthetic datasets. Their reported metrics demonstrate the pipelines on project data, not production or real-world model performance.