Abhiram Kandiyana

Graduate Research Associate @ University of South Florida

About

I am a PhD student in Computer Science at the University of South Florida, specializing in computer vision and deep learning, with a focus on using Vision Language Models (VLMs) for image classification and explanation. With a strong academic background and hands-on experience, I have developed frameworks that significantly improve image analysis efficiency (86%) while reducing the need for extensive ground truth data. My research has been presented at international conferences, and I have been recognized for my contributions to deep learning-based classification of biological images. I am skilled in deep learning, computer vision, and image processing techniques.

Country

United States

City

Tampa

Industry

Computer Software

Skill

CUDA, Dynamic Programming, Databases, Parallel Processing, C , Parallel Computing, dynamic program, Compute Unified Device Architecture (CUDA), Web Development, React.js, SQL, Predictive Modeling, OpenCV, Prompt Engineering, BERT (Language Model), Transformers, natural language processing, python, Transformer, Statistics

Experience

Graduate Research Associate

University of South Florida

2023-8 - Present · 3 yrs 2 mos

Tampa, Florida, United States

• Developed an active-learning framework achieving 91% image classification accuracy, equaling a trained CNN-based ensemble model while significantly reducing ground truth requirements (data and time) by 50x using VLMs and few-shot prompting techniques • Increased throughput efficiency by 86% by automating microscopy image classification using GPT-4 with minimal expert input • Automated the generation of detailed explanations for image analysis, providing interpretable insights that can be used for workforce training or fine-tuning vision-language models

Graduate Research Assistant

University of South Florida

2021-8 - 2023-7 · 2 yrs

Tampa, Florida, United States

• Developed an annotation framework that automatically counts and segments cells spanning multiple images while maintaining expert-level accuracy (>90%) using mean-shift clustering, a custom CNN(TinyCNN), and a fine-tuned version of Meta’s SAM. • Built supplementary annotation tools that run the backend’s semi-automated cell annotation framework to help experts annotate data 57% faster.

Software Engineering (Machine Learning)

ConceptWaves Software Solutions Pvt. Ltd.

2020 - 2021 · 1 yr

Hyderabad

Developed Machine learning frameworks to improve the accuracy, scalability and efficiency of Conceptwaves products.

Data Analyst

NeoAura

2019-5 - 2019-7 · 3 mos

visakhapatnam

• Improved shrimp production by 30% by predicting the ideal conditions for shrimp breeding using support vector machines and data analysis on the IOT sensors data from shrimp farming ponds.