Titinunt Kitrgungrotsakul.
I’m Titinunt Kitrungrotsakul, an AI researcher & data scientist. I build vision and multimodal systems for high-stakes domains — with strong evaluation and real demos.
Remote sensing imaging
Satellite CV at scale: segmentation/detection, domain shift, high-throughput inference pipelines.
- Slice-based evaluation (sensor/season/geography)
- Reproducible pipelines and dataset versioning
- Qualitative review workflows for failure modes
Medical imaging
Production-oriented segmentation/detection with reliability, uncertainty, clinician-facing outputs.
- Uncertainty triage and QA hooks
- Overlay visualization and failure-case dashboards
- Monitoring for drift and regressions
Data analysis
Turning ambiguous questions into metrics, experiments, and models stakeholders can trust.
- Decision-grade metrics with guardrails
- Experiment design and causal reasoning
- Dashboards and model cards
Selected publications
Keep this short and strong; put the full list on the Publications page.
- Extraction and Conversion of Carboxymethyl Cellulose from Okara Soybean Residue via Soda AQ Pulping: Integration of Predictive Models and Process ControlPreeyanuch Srichola, Titinunt Kitrungrotsakul, Kuntawit Witthayolankowit, Chaiyaporn Sampoompuang, Keowpetch Lobyaem, Prapakorn Khamphakun, Rawiwan TumthongPolymers 17(6): 777 • 2025
- A Cascade of 2.5D CNN and Bidirectional CLSTM Network for Mitotic Cell Detection in 4D Microscopy ImageTitinunt Kitrungrotsakul, Yutaro Iwamoto, Sari Ipponjima, Satoko Takemoto, Hideo Yokota, Tomomi Nemoto, Lanfen Lin, Ruofeng Tong, Hongjie Hu, Jingsong Li, Yen-Wei ChenIEEE/ACM Transactions on Computational Biology and Bioinformatics 18(5) • 2021
- Attention-RefNet: Interactive Attention Refinement Network for Infected Area Segmentation of COVID-19Titinunt Kitrungrotsakul, Yutaro Iwamoto, Hikari Jinbo, Sari Ipponjima, Tomomi Nemoto, Satoko Takemoto, Hideo Yokota, Lanfen Lin, Ruofeng Tong, Jingsong Li, Yen-Wei ChenIEEE Journal of Biomedical and Health Informatics 25(7) • 2021
- VesselNet: A deep convolutional neural network with multi pathways for robust hepatic vessel segmentationTitinunt Kitrungrotsakul, Xian-Hua Han, Amir Hossein Foruzan, Yutaro Iwamoto, Hideo Yokota, Tomomi Nemoto, Lanfen Lin, Ruofeng Tong, Jingsong Li, Yen-Wei ChenComputerized Medical Imaging and Graphics 75 • 2019
Showcase
Posters, architecture diagrams, qualitative grids, dashboards.
A two-stage training framework that first compresses a large "teacher" vision-language model into a smaller "student" using standard knowledge distillation, then further improves the student using reinforcement learning rewards.
This project presents VesselNet, a multi-pathway deep CNN for 3D hepatic vessel segmentation that classifies each voxel using three orthogonal 2D patches (sagittal/coronal/transverse) to better capture vessel structure in 3D.
A GraphRAG system designed for Vision-Language Models (VLMs) to answer employee questions from policy/legal documents and visually rich PDFs. It combines vector retrieval over multimodal chunks with a knowledge graph to improve scope correctness, reduce wrong-but-similar retrieval, and produce explainable answers with citations and reasoning paths.