Mohammadhossein Akbari Moafi
Hello
AI Researcher & Engineer with an M.Sc. in Computer Engineering from the University of Padova, specializing in machine learning and deep learning across diverse domains. My experience spans computer vision, recommendation systems, real-time AI applications, and emerging agentic systems, with growing expertise in large language models (LLMs), multimodal and graph-based learning, and a focus on knowledge graphs and semantic technologies for making research data findable and reusable.
Proficient in Python, C++, PyTorch, TensorFlow, and OpenCV, I combine strong software engineering practices — including DevOps/MLOps and scalable system design — with a passion for building robust AI systems that push the boundaries of efficiency and precision. My interests lie at the intersection of applied research and practical deployment of intelligent, agent-driven technologies.
Download my resumé as a PDF.
- Computer Vision
- Knowledge Graphs & Semantic Technologies
- LLMs & Agentic Systems
- Multimodal Learning
- MLOps & Scalable AI Systems
Experience
Responsibilities include:
- Developing and deploying ML, computer vision, and agentic LLM systems for real-world applications
- Experimenting with deep learning and multimodal architectures to improve model performance and reliability
- Building end-to-end AI and agent pipelines, from data preprocessing to evaluation, orchestration, and deployment• Developing and deploying ML, computer vision, and agentic LLM systems for real-world applications • Experimenting with deep learning and multimodal architectures to improve model performance and reliability • Building end-to-end AI and agent pipelines, from data preprocessing to evaluation, orchestration, and deployment
Responsibilities include:
- Designed an auto-labeling framework that ensembles UPerNet, OCRNet, and DeepLabV3+ using logit fusion and SDF-weighted integration for consistent semantic segmentation predictions.
- Refined ambiguous regions and boundries using zero-shot instance segmentation with the foundation model SAM, enabling label-agnostic fusion of semantic and instance outputs.
- Achieved a 5.8% mIoU improvement over the best individual model while reducing annotation time to ∼ 3 seconds per image, significantly minimizing manual effort.
Responsibilities included:
- Installed and configured servers, routers, switches, and operating systems (Linux & Windows).
- Supported setup and maintenance of security systems, including CCTV and burglar alarms.
- Assisted in troubleshooting and documenting network and system issues.
Education
- Thesis Title: Efficient Auto-Labeling for Enhanced Semantic Segmentation Annotation: An Ensemble-Based Framework with SAM Refinement
Accomplishments
