Mir Junaid
Scientific Computing Consultant (AI/HPC)
University of Duisburg-Essen
AI/ML Computing Expert specializing in High Performance Computing, Large Language Models, and Scientific Computing
mir.junaid@uni-due.de
About
I am a Scientific Computing Consultant at the University of Duisburg-Essen, specializing in High Performance Computing (HPC) and Artificial Intelligence. With a strong background in Machine Learning and Deep Learning, I focus on optimizing and scaling AI workloads for HPC environments, particularly Large Language Models (LLMs).
My expertise spans across running and optimizing AI applications on HPC clusters, managing GPU-accelerated computing workloads, and implementing efficient containerization solutions using NVIDIA NGC containers. I am passionate about bridging the gap between cutting-edge AI technologies and high-performance computing infrastructure.
Previously, I worked at Ecole centrale de Nantes and worked as research assistant at the Université de Technologie de Troyes, where I specialized in Graph Neural Networks and Geometric Deep Learning. I have extensive experience in both academic research and practical implementation of AI/ML solutions in HPC environments.
Experience
- Running Large Language Models (LLMs) on HPC clusters
- Porting and optimizing AI applications for HPC architectures
- Managing AI/ML workloads on HPC clusters
- Teaching and training students on effective HPC cluster utilization
- Deploying and managing NVIDIA NGC containers for GPU-accelerated high-performance computing and AI workloads
- Assisting researchers with porting and optimising AI applications to HPC architectures
- Handling AI/ML workloads on HPC
- Educating and training students on the effective utilisation of HPC clusters
- Contributing to the management of MesoNET HPC infrastructure
- Research and Application of Machine Learning: Graph Neural Networks, Geometric Deep Learning, and Graph Signal Processing
- Developed Attention-based GNNs inspired by Physics based Diffusion Models and Natural Language Processing
- Worked with Convolutional Neural Nets, Natural Language Processing, Graph-structured data, Network Analysis, Agricultural, and Weather Monitoring Data
- Classification of the pre-ictal and inter-ictal brain signals using Convolutional Neural Network
- Classification of fishes using Convolutional Neural Network
- Red Hat Certified Engineer (RHCE) and Red Hat Certified System Administrator (RHCSA)
Projects
GNN Visualization
Interactive visualization tool for Graph Neural Networks, demonstrating network architectures and data flow.
CNN Visualization
Visualization framework for Convolutional Neural Networks, helping understand internal network operations and feature maps.
Technical Skills
Programming Languages
- C++
- Python
- FORTRAN
HPC & Tools
- MPI
- OpenMP
- Git
- Linux
Areas of Expertise
- High Performance Computing
- Scientific Computing
- Numerical Methods
- Parallel Computing
Education
- Statistical/probabilistic ML (Bayesian ML, causal inference, approximate inference, modeling uncertainty)
- Advanced topics in representation learning (learning with little/no supervision, self-supervised learning, multi-modal representation learning)
- Graph Neural Networks, Geometrical Deep Learning, Computer Vision, Knowledge graphs, Knowledge-aware ML
- Applied ML in healthcare, drug discovery, and real-world settings (interpretability, ethics, ML Ops)
- Student Member (Technocrats-For-Electronics-Computer-Interaction-Society-IUST)
- Student Coordinator of ROBOMANIA-2013 organized by IIT Madras
- Student Coordinator of National Network Security Championship-2013 organized by IIT Delhi