Multimodal Video Understanding
Designing chaptering and summarization systems that align spoken, visual, and on-screen textual signals from presentation-style videos.
Applied Scientist / ML Engineer Track
Building end-to-end ML systems for computer vision, multimodal AI, and video understanding.
I am a PhD Candidate in Computer Science at the University of Houston. My work focuses on computer vision, multimodal AI, and vision-language systems for educational video understanding, multimodal summarization, object detection, and robust evaluation pipelines.
- Research Areas CV, Multimodal AI, VLMs, LLM Evaluation
- Core Stack Python, PyTorch, OpenCV, Transformers, Docker
- Current Goal Applied Scientist and Machine Learning Engineer roles
About
Research-trained engineer focused on practical ML systems
My research sits at the intersection of computer vision, multimodal AI, and machine learning systems. I design pipelines that combine ASR transcripts, OCR text, slide visuals, and detected visual objects to improve navigation and review for long educational videos.
Beyond model development, I care about modular pipeline design, experiment reproducibility, evaluation rigor, and translating research ideas into systems that are easier to use, benchmark, and extend.
Focus Areas
What I work on
Vision-Language & LLM Evaluation
Building automated workflows to measure faithfulness, relevance, coherence, and ranking quality for generative multimodal systems.
Reproducible ML Pipelines
Creating modular, configuration-driven experimentation frameworks for training, inference, debugging, and iterative model development.
Featured Work
Projects and repositories
Flagship Project
Multimodal AI
video-chaptering
End-to-end multimodal pipeline for educational video chaptering and summary generation using ASR, OCR, slide visuals, and supporting visual evidence.
- Modular stage-wise workflow for prepare, keyframes, modalities, and chaptering
- Designed for evidence alignment, summary generation, and iterative experimentation
- Focused on practical review and navigation utility for long lecture videos
Video Understanding
LLMs
LLM-Based Video Temporal Segmentation
Slide-anchored segmentation workflow for dividing long educational videos into topically coherent chapters with validation and retry logic.
- Combines transcript and visual structure to improve segmentation quality
- Designed for robust output correction and quality control
- Supports downstream chapter-level summarization workflows
Computer Vision
PyTorch
SOTA Object Detection Lab
Reusable training and evaluation framework for object detection experiments with configuration-driven workflows and rapid iteration.
- Supports shared train, evaluate, and predict workflows
- Structured for reproducibility and model comparison
- Useful for custom dataset fine-tuning and experimentation
Applied ML
Multimodal Systems
Real Estate Listing Optimization
Pipeline for collecting, structuring, and analyzing listing data with geospatial filtering and multimodal recommendation workflows.
- Built for large-scale listing extraction and comparative analysis
- Combines structured data pipelines with multimodal model outputs
- Targets practical pricing and content optimization use cases
Technical Stack
Tools and technologies
Languages
Python, SQL, C++, MATLAB, R, PHP, JavaScript, HTML
ML / CV / Multimodal
PyTorch, TensorFlow, Hugging Face Transformers, OpenCV, Whisper, DeepEval
Tools / Platforms
Docker, Linux, Git, Hydra, Slurm, MLflow, Weights & Biases, AWS EC2, AWS SageMaker
Selected Research
Publications and research themes
Educational video understanding and summarization
Research on presentation-style educational videos, chapter-based summarization, and multimodal evidence selection from transcripts, OCR, slides, and detected visual objects.
Visual content detection in lecture videos
Work on detecting pedagogically meaningful objects such as charts, diagrams, tables, and text regions to complement generated summaries.
Add direct paper links, Google Scholar links, or project pages here once you decide which 2–4 items you want to feature publicly.
Contact
Let’s connect
I am interested in Applied Scientist and Machine Learning Engineer opportunities in computer vision, multimodal AI, and video understanding.