Interests
- Multimodal representation learning (text–image–video)
- Evaluation and alignment of vision–language models
- Retrieval-augmented generation and video summarization
- Safety, brand suitability, and content classification
I build and evaluate multimodal AI systems.
I work on vision–language models, large-scale video understanding, and practical MLOps for deploying these systems in real products. I’m especially interested in retrieval-augmented pipelines, representation learning, and evaluation for safety and robustness.
- Multimodal LLMs & video understanding
- Large-scale evaluation & content safety
- End-to-end pipelines from research to deployment
Current focus
- Designing and evaluating video search & summarization pipelines.
- Fine-tuning vision–language models for safety and classification.
- Building tools to read and analyze research papers more efficiently.
About
I am a machine learning researcher working on multimodal AI and real-world deployment of vision–language systems. My recent work focuses on video search and summarization, content understanding, and evaluation pipelines that connect research models to production constraints.
Broadly, I enjoy working at the boundary between research ideas and systems that actually ship: designing models and representations, and then building the data, training, and inference infrastructure needed to make them useful in practice.
Research & Publications
I am currently organizing my publications, patents, and project notes. A more detailed, paper-style list will appear here soon.
Selected Publications
-
MMTF: Multi-Modal Temporal Fusion for Commonsense Video Question Answering
2023 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW), 2023
-
Plant disease detection in imbalanced datasets using efficient convolutional neural networks with stepwise transfer learning
IEEE Access, 9:140565–140580, 2021
-
Image classification based on automatic neural architecture search using binary crow search algorithm
IEEE Access, 8:189891–189912, 2020
-
Heterogeneous visible-thermal and visible-infrared face recognition using cross-modality discriminator network and unit-class loss
Computational Intelligence and Neuroscience, 2022
Patents
-
Electronic device and operation method for multimodal temporal-axis fusion AI models
U.S. Patent No. 12,299,967 (Granted May 2025) and Korean Patent No. 10-2812533 (Granted May 2025)
Co-inventor -
Electronic device for multi-modal temporal-axis fusion artificial intelligence models and operation method thereof
Korean Patent No. 10-2721151 (Granted Oct 2024)
Co-inventor -
Apparatus and method for generating a learning model for classifying images using transfer learning
Korean Patent No. 10-2641533 (Granted Feb 2024)
Co-inventor -
Method and apparatus for automatic design of artificial neural network structure based on crow search algorithm
Korean Patent No. 10-2694148 (Granted Aug 2024)
Co-inventor
Selected themes
- Building evaluation pipelines for large-scale video catalogs
- Leveraging LLMs and VLMs for automatic labeling and curation
- Designing long-form understanding tasks over video segments
Projects
Below are a few representative areas I’ve worked on recently. Some are research-oriented; others are closer to production ML engineering.
Large-Scale Video Understanding Pipeline
End-to-end pipeline for semantic video analysis and retrieval using multimodal embeddings, automatic speech recognition, and LLM-based content understanding. Developed novel fusion techniques for combining visual, audio, and textual features at scale.
Multimodal AI Safety & Classification
Fine-tuning and evaluation frameworks for vision-language models in content classification tasks. Developed interpretable multi-label classification systems with emphasis on robustness, fairness, and explainable AI outputs.
Academic Research Acceleration Tools
AI-powered tools for efficient academic paper analysis, including automated highlighting, semantic note-taking, and LLM-assisted literature review workflows. Designed to accelerate research discovery and knowledge synthesis.
Experience (short)
A detailed CV is available on request; this is a compact snapshot of my recent work.
-
Machine Learning Researcher
Working on video understanding, multimodal retrieval, and large-scale evaluation frameworks that connect LLMs/VLMs to real production workloads.
-
PhD in Computer Science
Research on model design and training methods with a focus on reducing human effort and deploying practical systems.
Contact
I’m open to conversations about research collaboration, postdoctoral opportunities, applied ML projects, and practical deployment of multimodal systems.
Email: ahmadmobeen24@gmail.com