“Dream is not what you see in sleep;
It is that which doesn't let you sleep.”
- A.P.J. Abdul Kalam
“Dream is not what you see in sleep;
It is that which doesn't let you sleep.”
- A.P.J. Abdul Kalam
Hi, I'm Johirul Islam, currently working as a doctoral researcher at the University of Oulu, Finland. As a doctoral researcher, by accommodating the extreme-edge local computing cluster, I am trying to reshape the traditional 2-tier edge-cloud into a 3-tier computing architecture for building a smart hospital environment.
During my doctoral research, I have had the chance to play with various RTOS-based IoT devices (e.g., RPi Pico, ESP32, Arduino Nano, etc.) and GPOS-based IoT devices (e.g., RPi 3, RPi 4, RPi 5, etc.) to build and deploy the required services. I also utilized Nvidia GPU resources (e.g., Jetson Xavier NX Dev Kit & Clara AGX) for developing demos and PoCs for medical-use-case services. The services are deployed into a 3-tier architecture, considering extreme-edge nodes as tier-1 while MEC-based nodes and Cloud nodes are considered tier-2 and tier-3, respectively.
I build things from scratch with my strong knowledge of Linux, Docker, Kubernetes, and cloud-native tools. I don't just deploy the required services; I set up bare-metal Kubernetes clusters, write custom automation, and dig into the underlying system until things work the way they should. That same instinct is now driving my transition into AI/ML—building pipelines, understanding the stack end-to-end, and bridging the gap between infrastructure and intelligent systems.
Currently, I am expanding my knowledge into generative AI while continuing to deepen my infrastructure expertise—because the best AI/ML systems are only as good as the infrastructure they run on.
If you are interested in knowing about me, please leave a message on LinkedIn and visit my Projects.
Programming & Scripting: C, C++, Java, Python, Bash (Linux).
AI / ML Tools and Framework: Pytorch, Tensorflow, LLM, SLM, Ollama.
Version Control: Git, GitHub, Bitbucket & GitLab.
Virtualization & Orchestration: VMs (VirtualBox & VMware), Containerization (Docker & Kubernetes).
CI/CD & Automation: GitLab, Ansible, and N8N.
Monitoring & Logging: TIG (Telegraph, InfluxDB & Grafana) stack.
Embedded systems: ESP32, Arduino Uno, Raspberry Pi Pico (RPi Pico), and Raspberry Pi (3|4|5).
GPU systems: Nvidia Jetson Xavier NX Dev Kit, Nvidia Clara AGX, and Nvidia Blackwell.
🎓 Doctoral Researcher, University of Oulu, Finland (Feb 2021– Present)
Objectives: How lightweight services can be deployed into constrained medical IoT devices, ensuring scalability, reliability, and interoperability across heterogeneous platforms with optimal use of resources.
Implementation: Nanoservices are deployed into SBCs, e.g., Raspberry Pis & Nvidia Jetson Xaviers & Nvidia Clara AGX (for GPU applications), focusing on the OSI 7th (application) layer with CoAP and MQTT protocols.
Results: article 1 (dynamic service deployment) and article 2 (GPU utilization through Docker container).
🎓 Research Assistant, University of Oulu, Finland (Dec 2018 – Jan 2021)
💻 Software Engineer, Rokomari.com, Dhaka, Bangladesh (Jun 2015 – Jul 2016)
Role: Backend software developer.
Technologies: Java, Scala, PHP, Maven, Spring (framework), MySQL, PostgreSQL, etc.
💻 Junior Software Engineer, DuttaSoft, Dhaka, Bangladesh (Oct 2014 – May 2015)
Role: worked as a web developer.
Technologies: PHP, CodeIgniter (framework), MySQL, etc.
Objectives: How can services be deployed dynamically into a heterogeneous IoT network based on the service requirement and resource availability?
Implementation: Services deployed into 5 non-uniform Raspberry Pi (RPi) nodes with 2 manager nodes with Docker and Docker Swarm.
Publication: Results related to the experiment are published in the IEEE Access Journal.
Objectives: Measuring component-wise energy consumption of 5G private networks.
Implementation: Energy consumption of Allbesmart's OAIBox (core + gNB) and Quectel modem (UE) is monitored with a Netio Powerbox (a collaboration project done with VTT & University of Oulu).
Publication: Results related to the experiment are published in the ACM Digital Library.