I build production-grade agentic AI systems — multi-agent orchestration with LangGraph, retrieval-augmented LLMs, voice-first agents, and computer vision pipelines. Currently engineering a workflow orchestration platform at FlowGenX AI and shipping the open-source agent observability framework Retune.
I'm an AI/ML engineer focused on the agentic side of modern AI — designing systems where multiple LLM-powered agents coordinate, call tools, and recover from failure inside production workflows. Over the last few years I've shipped multi-agent ERP automation, LangGraph-based orchestration platforms, fine-tuned LLMs for domain-specific RAG, and computer-vision pipelines that run in real time on edge hardware.
Before going full-time on AI, I spent years on backend systems (FastAPI, Celery, .NET Core, microservices) and Android. That foundation shapes how I build ML — with careful attention to latency, observability, queueing, and the messy realities of getting a model from a notebook into a service that someone depends on. My open-source work (Retune) and a research record in malware detection, blockchain bibliometrics, and Bangla NLP keep me close to both ends of the stack.
ESAP AI (Empowering Energy)
NybSys Inc.
Pro Data Ltd.
A selection of work I'm most proud of — from production agentic platforms to launched consumer products.
An automation framework for designing, orchestrating and deploying AI agents and workflows. Node + edge graph runtime with ReAct agents, supervisor agents, routers, loops, triggers, MCP tools, and enterprise app connectors. Comparable in scope to n8n, MuleSoft and Boomi — but agent-native.
A complete dev-tool for AI agents: observe actions, tool calls and I/O; evaluate against criteria; optimize agent behavior — internally powered by a Deep Agent architecture. Ships with an SDK and a dashboard, free and open for contribution.
A consumer ride-sharing product I designed and launched for Bangladesh. Riders post a trip with pickup and destination; drivers (cars, bikes, CNG, trucks) bid live; the rider negotiates and books. Scalable architecture supporting both web and (soon) native mobile clients.
Fully automated ERP with HR and smart price-quotation modules. A Coordinator Agent orchestrates voice-input specialist agents using LangGraph, LangChain and Phidata, with web-scraping pipelines for live vendor pricing.
A YOLOv11-seg model fine-tuned on custom-annotated room imagery to segment walls, floors, ceilings and furniture, paired with AR-based color recommendation logic. Annotation pipeline built on Label Studio.
Llama 3.2-3B with LangChain prompting and FAISS retrieval over the Quran (translation + explanation), with sentence-transformer verse retrieval.
Fine-tuned YOLOv8 for human detection with OpenCV optimisations (counting, saving, frame compression) and FastAPI real-time inference.
Custom agentic chatbot fine-tuned on company policy data, with FAISS retrieval and a FastAPI inference pipeline serving customer support.
Optimized scraping for product price & vendor data using Google Search API, DuckDuckGo, Selenium, Playwright. Embedded and filtered via Ollama OSS LLMs.
Fine-tuned runwayml-stable-diffusion-v1-5 on a custom fashion dataset for image-to-image generation — a learning-focused exploration of diffusion training.
BERT-based Text2Text generation model fine-tuned on a medical symptoms-to-disease dataset; deployed with a Gradio QA interface on Hugging Face Spaces.
Peer-reviewed work in machine learning, security, and NLP.
Mahmud, A., Hossain, M. S., Meya, M. R., Didar, D. I., Islam, H., Dhar, P., Akter, P. — Lecture Notes in Networks and Systems, pp. 409–424.
DOI →Hossain, M. S., & Haque, A. B. — Intl. Conf. on Machine Intelligence and Emerging Technologies, Springer Nature, pp. 348–362.
Google Scholar →Hossain, M. S., & Riaz, M. H. — Intelligent Computing & Optimization (ICO 2021), Springer International Publishing, pp. 277–287.
Google Scholar →Noakhali Science and Technology University · CGPA 3.30 / 4.00
Thesis: Android Malware Detection System — A Machine Learning and Deep Learning Based Multilayered Approach.
Undergraduate Research Assistant (Mar 2020 – Jan 2021) under Md Hasnat Riaz, Assistant Professor, Dept. of CSTE.
Feni Govt. College · GPA 4.83 / 5.00
Central High School · GPA 5.00 / 5.00
Open to Senior AI/ML Engineer roles, agentic-systems consulting, and research collaboration.