About
AI Engineer and Game Developer specializing in production-grade RAG systems, LLM adaptation, and real-time Unreal Engine 5 gameplay mechanics. Currently working as a Junior AI Engineer at Softvence Delta.
Skilled in Python and C++, I build scalable workflows using LangGraph, LangChain, and Hugging Face. My practice covers the full lifecycle from LLMOps and MLOps to interactive game systems.
I bridge the gap between machine learning depth and robust gameplay engineering to deliver scalable, high-performance architectures.
Projects
GPT-2 From Scratch
Built a GPT-2 style transformer architecture from scratch using PyTorch. Implemented token embeddings, positional embeddings, multi-head self-attention, feed-forward networks, layer normalization, training pipeline and text generation.
ReAct Agent
Built a LangGraph-powered AI Agent implementing the Reason -> Act -> Observe workflow with MCP tools, PostgreSQL integration, FastAPI backend and Docker deployment.
Bangla Sign Language Recognition
Developed a real-time Bangla Sign Language recognition system using YOLO, achieving 98% F1 score across 36 alphabet classes.
Skin Cancer Detection
Built a full-stack skin lesion classification platform using YOLOv12 and FastAPI with real-time inference capabilities and 91% accuracy.
AI Boss Fight System
Created an Unreal Engine 5 AI-driven boss system featuring Behavior Trees, adaptive combat behaviors, AOE attacks, spell casting and multi-phase encounters.
Experience
Junior AI Engineer
Softvence DeltaBuilt production RAG systems, LangGraph workflows, LLM adaptation pipelines, MLOps automation, and enterprise AI solutions.