One-Shot Adaptive Segmentation for Scientific Images
Training-free DINOv3–SAM adaptation from one annotated reference image.
AI researcher · Purdue University
Tejaswi
Panchagnula
I build adaptive AI systems across computer vision, language models, and agentic architectures.
Graduate Research Assistant, VIPER Lab
Advised by Dr. Fengqing Maggie Zhu ↗
01 / Publications
Selected research
arXiv ↗
Inference-Time Mitigation of Adversarial Political Bias in Large Language Models
Self-correction that improves political neutrality without additional training.
Foraging with the Eyes: Dynamics in Human Visual Gaze and Deep Predictive Modeling
A large-scale study of Lévy-walk dynamics in human visual search.
02 / Selected work
Built and tested
01
Live product
Agentic AI PDF Editor
A conversational platform that combines an editing agent, computer vision, and OCR to modify PDF forms from natural-language instructions.
Gemini 2.5 · React · FastAPI · YOLO · OCR
02
1st place · Purdue Chips & AI Hackathon
Compute Commons
A shared AI-training network that coordinates learning across idle, heterogeneous community devices.
PyTorch · MLX · WebSockets · Streamlit
03
Research code
Gaze Analysis
Large-scale gaze dynamics and fixation prediction with more than four million recorded points.
PyTorch · Computer Vision · IIT Madras
03 / About
Research with a builder’s instinct.
I’m an M.S. student in Electrical and Computer Engineering at Purdue University and a Graduate Research Assistant in the VIPER Lab.
Previously, I was a Research Associate at the Walmart Center for Tech Excellence and a Research Intern at SENAI Lab, IIT Madras.
Python · PyTorch · vLLM · LangChain · OpenCV · AWS