Safal Thapaliya

Email | Google Scholar | LinkedIn | GitHub | CV

Hello there! I’m Safal Thapaliya, a second-year PhD student in Computer Science and Engineering at the University of Connecticut, Storrs (School of Computing), advised by Prof. Chuxu Zhang.

Research interests and focus:

Large language models (LLMs)
Resource-efficient models
Trustworthiness in graph neural networks (GNNs)
AI for healthcare

Background:

B.E. in Computer Engineering, IOE, Pulchowk Campus (2022)
Research Assistant at NAAMII (2022–2024), supervised by Dr. Bishesh Khanal; collaborated with KIAS and Dr. Taman Upadhaya
- Worked on vision-language models (VLMs), medical image segmentation, and object detection
- Built healthcare AI prototypes and deployment-ready pipelines

Other:

Founding member and developer of Clamphook (2019-2022)
Internship at Logpoint’s Support and Solutions Department (2022); Linux Server Administration, Distributed Log Handling

Publications

* indicates equal contribution

Semantic Refinement with LLMs for Graph Representations

Safal Thapaliya*, Zehong Wang*, Jiazheng Li, Ziming Li, Yanfang Ye, Chuxu Zhang

(2025)

arXiv pre-print

PDF

TuneVLSeg: Prompt Tuning Benchmark for Vision-Language Segmentation Models

Rabin Adhikari, Safal Thapaliya, Manish Dhakal, Bishesh Khanal

(2024)

Asian Conference on Computer Vision (ACCV 2024)

Oral Presentation

PDF Code

Vlsm-adapter: Finetuning vision-language segmentation efficiently with lightweight blocks

Manish Dhakal, Rabin Adhikari, Safal Thapaliya, Bishesh Khanal

(2024)

International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI 2024)

PDF Code

Exploring Transfer Learning in Medical Image Segmentation using VLMs

Kanchan Poudel*, Manish Dhakal*, Prasiddha Bhandari*, Rabin Adhikari*, Safal Thapaliya*, Bishesh Khanal

(2024)

Medical Imaging in Deep Learning 2024

Oral Presentation

PDF Code

For a complete list of publications, see my google scholar.

Projects

AI Assisted Smartphone Microscopy (2024)

Built and deployed a role-based web platform that enabled multi-site collection and annotation of 400K+ microscopic images using Django; benchmarked state-of-the-art object detection models for parasite detection on these images.

Lower Limb Angle Measurement for Corrective Osteotomy (2023)

Built and deployed a CT-scan-based pipeline using U-Net/nnU-Net for bone segmentation and landmark detection to automatically compute lower-limb alignment angles, containerized with Docker and served via Flask for clinical-style use.

Public Discourse Analysis System (2022)

Architectured a real-time tweet analysis platform using Flask and MongoDB; containerized the application with Docker for scalable deployment.