LuxenAI/arbor
A framework for structuring, visualizing, and orchestrating agent behavior in a more interpretable tree-based form.
Last updated: July 2026
Ganesh Talluri
Hi! I'm Ganesh Talluri and I'm currently based in Phoenix, AZ. I'm currently working on computer vision models for Axon, and I am also an ML student researcher at the SoftBank Group, advised by Hiroki Naganuma.
My research interests span foundation models, representation learning, distribution shift, and machine learning for scientific discovery. I have worked on foundation models, spectral optimizers, distributed training, and machine learning under distributional shift. More recently, I have become interested in low-level systems and kernel optimization for efficient large-scale model training and inference.
Most recently, I was a Research Assistant at Midwestern University, advised by Prof. Andrew Lee, a Venture Fellow at Harvard University and a Mercatus Fellow at George Mason University. I was also an Associate Scholar at the Masason Foundation and the co-founder of Luxen LLC.
Feel free to reach out if you are interested in my work.
Education
High School Diploma
- Cumulative GPA: 4.7/5.0
- 15 APs, 1510 SAT
- Attended the #1 ranked public high school in the U.S. (U.S. News 2024)
- National Merit Scholarship Commended Scholar, AP Scholar with Distinction x2, CollegeBoard National Recognition Program
- President of Science Olympiad, Captain of Academic Decathlon, and President of USABO Club
Mercatus Fellow
- Selected as 1 of 48 fellows for a competitive Mercatus Center fellowship with a less than 10% acceptance rate.
- Received a stipend to study political economy, institutional analysis, and governance with George Mason University scholars.
- Applied economic frameworks to questions of entrepreneurship, markets, public policy, and institutional design.
Experience
Machine Learning Engineering Intern
Cofounder
- Founded an applied-AI lab focused on autonomous agent development.
- Built lightweight orchestration layers: LuxenAI/arbor for agent behavior and LuxenAI/regimen for small-model workflows.
- Raised $48K in grant funding and equipment from 1517 Fund, SoftBank Group, and BizWorld.
- Reached 5K+ active users and over 1.5M media views.
Machine Learning Student Researcher
- Conducted a large-scale empirical study of 40+ deep learning generalization measures across 10,000+ training configurations, identifying which metrics reliably predict model performance under distribution shift and out-of-distribution settings.
- Found a geometric mismatch in low-rank spectral optimizer which once corrected achieved the convergence of full rank spectral optimizers.
- Worked on spectral optimizers, distributed training regimes, and out of distribution generalization measures.
- Work accepted to ICML 2026 CoLoRAI, pending review at NeurIPS 2026.
Research Assistant
- Co-developed BONe deep learning software modules for training, prediction, and evaluation of microCT bone segmentation models.
- Evaluated model robustness using 5-fold cross-validation, 3 random seeds per fold, and 30 architecture/backbone/patch-size benchmarking experiments.
- Poster presentation at Anatomy Connected 2026.
Publications
1. Tatsuhiro Nakamori, Laura Gomezjurado Gonzalez, Ganesh Talluri, Ansh Tiwari, Hideyuki Kawashima, Ioannis Mitliagkas, Guillaume Rabusseau, Hiroki Naganuma.
"Orth-Dion: Eliminating Geometric Mismatch in Distributed Low-Rank Spectral Optimization."
ICML'26 CoLoRAI Workshop (05/2026).
[Paper]
2. Ganesh Talluri.
"Glyde: A Domain-Aware, Topology-Biased Glycan Language Model for Viral Receptor Binding."
SSRN (04/2026).
[Paper]
3. Sora Nakai, Youssef Fadhloun, Kacem Mathlouthi, Kotaro Yoshida, Ganesh Talluri, Hiroki Naganuma, Ioannis Mitliagkas.
"Revisiting Generalization Measures Beyond IID: An Empirical Study under Distributional Shift."
arXiv (02/2026).
[Paper]
4. Andrew H. Lee, Ganesh Talluri, Manan Damani, Brandon Vera Covarrubias, Helena Hanna, Julian M. Moore, Jacob Baradarian, Jeremy Chavez, Michael Molgaard, Beau Nielson, Kalah Walden, Thomas L. Broderick, Layla Al-Nakkash.
"New software and 2D models to segment bone in micro-CT scans."
Frontiers in Bioinformatics (01/2026).
[Paper]
5. Ganesh Talluri, Ashwin Rokkam.
"A Novel Multi-Omics Deep Learning Framework for Spatiotemporal Cerebral Cortex Localization & Expression."
SSRN (05/2025).
[Paper]
6. Ganesh Talluri, Ashwin Rokkam.
"Sculpt: Novel Multi-Omics AI for Neuronal Modeling."
medRxiv (04/2025).
[Paper]
Projects
Interactive Transformer
An interactive visualization project for understanding transformer internals, attention behavior, and model structure more intuitively.
More projects on my GitHub.
Awards, Fellowships & Funding
Fellowships & Selective Programs
- Masason Foundation Scholar (<5%) — Tuition and Board Scholarship 06/2025
- Harvard Ventures-TECH Winter Fellow (<0.5%) — Entrepreneurship and Venture Capital Fellowship 12/2024
- ZFellows Finalist (<10%) — 2× Finalist 12/2024 & 12/2025
Honors
- Academic Decathlon (<0.1%) — Rank #1 at Individual Nationals Scholastic Division 05/2025
- Arizona Future Innovator of the Year - Research Award, Honorable Mention 11/2025
- SoftBank — Research Grant ($32,000) 01/2026
- Gifted Talented (<3%) — 20 Under 20 Awardee. $2,000 Support Grant 04/2026
- USA Biology Olympiad (USABO) — 2x Honorable Mention 02/2024 & 02/2025
- Academic Decathlon — Rank #2 in Regional Competition. $10,000 Scholarship 05/2025
- Arizona Elks MVS Award — $2,000 Merit-based Scholarship 05/2026
- Arizona Science & Engineering Fair — 2x 2nd Place Grand Award, Computational Biology 03/2025 & 03/2026
- International Biology Battle — 1st Place Team Worldwide (1/150). Top 30 individually (30/400) 04/2025
Miscellaneous Awards/Distinctions
- 1517 Fund — Startup Grant 04/2025
- BizWorld YES! Accelerator — Startup Grant 05/2025
- Y Combinator Startup School — Startup Credits 04/2026
- Network School — Startup Community 04/2025
- Sigma Xi — Research Honor Society Member 04/2026
- Arizona State University Grand Challenges Scholar — Research Grant 03/2026
- Microsoft for Startups — Startup Credits 05/2024
Skills
Programming & Systems
Python, Java, JavaScript, CUDA, HTML/CSS
ML & Training Infrastructure
PyTorch, JAX, Transformer Architectures, LLM Training, FSDP/ZeRO
Optimization & Theory
Optimizer efficiency, Numerical Linear Algebra in ML, Communication Bottlenecks & Quantization (GPTQ, QAT), Probability & Statistics for Machine Learning