Anagha Joshi

I am ML researcher and software engineer with over eight years of experience building large-scale applications and practical AI systems. I hold an M.S. in Computer Science from the University of Georgia, Athens, USA, where I focused on interpretable AI for image classification.

Based in Austin, Texas, I work as a Software Development Engineer at Amazon, where I design scalable APIs, automate complex systems, and write clear technical documentation that keeps teams aligned. I am now expanding my research into projects that connect my industry background with real-world impact, especially in explainable AI and AI systems that help organizations make better decisions with limited resources.

My goal is to help build technology that is robust, understandable, and accessible — whether that means designing multi-agent AI systems, deploying scalable APIs, or improving trust in AI-powered recommendations for everyday businesses.

Email / CV / LinkedIn / GitHub

Research

My research bridges software engineering and practical AI. I have developed AI-driven program analysis tools using Boolean tensor models for syntax error repair and explored core-set active learning with dimensionality reduction to improve text classification on real-world datasets.

During my master’s, I designed TPNet, an interpretable transformer-based image classifier that shows how deep learning models can remain transparent and auditable. I also built a deployed web application that automated grading tasks, combining practical software design with applied machine learning.

I am motivated by research that tackles real-world challenges at the intersection of technology, environment, and human well-being. I am interested in building AI systems that stay understandable and make a measurable difference where they are needed most.

Publications

A. N. Joshi, "Revitalizing Language Processing with Intelligent Syntax Repair and Boolean Tensor Completion," 2024 IEEE International Conference on Big Data (BigData), Washington, DC, USA, 2024, pp. 5361–5368.
[IEEE Xplore] · doi:10.1109/BigData62323.2024.10825203

Anagha Joshi. 2025. "Enhancing Core-Set Active Learning: Unlocking New Frontiers in Text Classification." In Proceedings of the 2024 8th International Conference on Natural Language Processing and Information Retrieval (NLPIR '24). Association for Computing Machinery, New York, NY, USA, 62–68.
[ACM Digital Library] · doi:10.1145/3711542.3711606

A. Joshi, “Interpretable Image Classification,” M.S. Thesis, University of Georgia, 2021.
[UGA Repository]

A. Joshi, M. V. Kulkarni, "Analysis of Fractal Image Compression Using Quadtree Partitioning," Bulletin of Marine Science & Technology, Vol. 10, 2015.

Projects

SynSculpt – Syntax Error Repair Model: I developed a method using boolean tensor completion to suggest and rank fixes for programming language syntax errors, achieving ~90% accuracy in identifying valid code repairs.

Core-Set Active Learning Enhancements: I applied dimensionality reduction (t-SNE) within core-set active learning for text classification, improving F₁ scores by 2–3% and highlighting cases where core-set selection underperforms random sampling.

TPNet (Interpretable Image Classifier): I designed a transformer-based image classification model that provides visual explanations for its predictions, reducing model complexity while maintaining accuracy (M.S. thesis project).

Autograding Web Application: I built a Django web app for automated grading of assignments, which automated 100% of grading tasks and reduced manual grading time by 90%, deployed for an educational research lab.

Research Blog

(Coming soon) I plan to share articles and insights about my research, covering topics in software engineering, NLP, and machine learning. Stay tuned for updates!

Artwork

I enjoy creative pursuits outside of computing. I will be showcasing some of my artwork and creative projects here soon.

Sports

I have a passion for sports and fitness. I am a powerlifting athlete and recently achieved a personal record deadlift of 200 lbs (2023). I also hold a Black Belt (Dan) in Muay Thai (earned in 2018). Staying active through strength training and martial arts helps me build discipline and energy for my research.

Contact

Feel free to reach out to me via email, or connect with me on LinkedIn and GitHub. You can also download my full CV here.

Based on the source code from jonbarron.info.