Hi, I’m Mahbub Islam Mahim, a Software Engineer at Samsung Research, where I work on knowledge graphs and reasoning systems, developing a Knowledge Graph build tool using the RDFox engine in collaboration with Oxford Semantic Technologies , UK.
I completed my BSc and MSc in Computer Science & Engineering at Jahangirnagar University, and was awarded the National Science and Technology Fellowship during my studies. In February 2025, one of my research projects at Samsung earned an A1-grade patent, which I’m especially proud of.
Experience
Senior Software Engineer
Following Samsung Electronics' acquisition of Oxford Semantic Technologies in 2024, I now work directly within Samsung's integrated AI ecosystem on intelligent knowledge graph systems that help organizations make sense of their data. I contribute to building semantic platforms that transform scattered information into structured, queryable knowledge, enabling better decision-making and insights across enterprise environments. I am currently leading the development of a key module within the platform in collaboration with Oxford Semantic Technologies, using RDFox's cutting-edge reasoning capabilities to power Samsung's next-generation knowledge-based AI applications.
Patent
Method for Enhancing Wide-Angle Images by Integrating Multiple Fields of View from Different Lenses
A novel method that enhances wide-angle images by integrating multiple fields of view from different lenses, providing superior image quality and expanded visual coverage. This innovation was graded A1 by Samsung Research on February 20, 2025, representing the highest level of technical excellence and commercial potential.
Research & Publications
Divide and Conquer: A Two-Stage Cascaded Framework with K-Fold Ensembling for Multi-Label Bangla Hate Speech Classification
Designed a label-wise cascaded pipeline where each label is first classified using a binary model and then refined with a multiclass model, with k-fold cross-validation and intra-label ensembling. Became CHAMPION in the workshop contest and currently writing for BEST PAPER award.
Optimizing the Fine‑Tuning Process of Large Language Models
A comprehensive comparison of a memory-efficient fine-tuning method that combines quantization and LoRA on LLaMA-2 7B using the CodeAlpaca-20k dataset, reducing GPU memory usage from 112 GB to 10.8 GB without sacrificing performance, thereby enabling low-resource LLM deployment in real-world applications such as coding assistants.
Enhancing Wide‑Angle Image Using Narrow‑Angle View of the Same Scene
Developed a GAN-based approach to enhance wide-angle images by fusing fine details from co-captured narrow-angle views, leveraging attention-based feature integration to improve visual fidelity.
Awards & Recognition
Icon Engineer
Icon Engineer
Recognition in Innovation
Professional Software Coding Certification
Bronze Medal
Programming Contests
Former competitive programmer with strong algorithmic problem-solving skills. Participated in major programming contests including ICPC, National Programming Contest, and various online platforms during university years.
ICPC
3-time participant (2021-24)
NCPC 2023
NCPC finalist
Programming Statistics
Activities
My Ongoing Journey Through Work, Life, Learning, and the Experiences That Shape Who I Am!
Patent Experience Sharing Session at SRBD
April 15, 2025
Conducted a session on patent strategy in front of engineers at SRBD, specifically sharing my A1 patent journey experience. Demonstrated the complete process from ideation to patent filing, including best practices for intellectual property protection and strategies for successful patent applications.