cv
Basics
| Name | Karthik Viswanathan |
| Label | Researcher |
| vkarthik095@gmail.com | |
| Phone | (31) 684103282 |
| Url | https://karthikviswanathn.github.io/ |
| Summary | Topological Data Analysis for Cosmology \( \to \) Interpretability in LLMs |
Education
Work
-
2017.05 - 2019.05 Surveillance Analyst
Goldman Sachs, Bangalore, India
Developed ML models for anomaly detection in financial data. Manager: Prof. Howard Karloff
-
2016.05 - 2016.07 Summer Intern
Goldman Sachs, Bangalore, India
Implemented fast Personalized PageRank in a MapReduce framework. Mentor: Dr. Koushik Balasubramanian
Publications
-
2025.07.01 Probing Geometry of Next Token Prediction Using Cumulant Expansion of the Softmax Entropy
ICML 2025: Workshop on High-dimensional Learning Dynamics
We introduce a cumulant-based framework to analyze how transformer models capture higher-order statistics during next-token prediction.
-
2025.01.17 The Geometry of Tokens in Internal Representations of Large Language Models
arXiv
We investigate the relationship between the geometry of token embeddings and their role in next token prediction.
-
2024.10.14 Persistent Topological Features in Large Language Models
ICML 2025
We use zigzag persistence to describe data undergoing dynamic transformations across layers of LLMs.
-
2024.09.18 Cosmology with Persistent Homology: a Fisher Forecast
Journal of Cosmology and Astroparticle Physics
We apply persistent homology to dark matter halo catalogs and find tighter constraints on cosmological parameters.
Research Programs and Workshops
- 2026.01.01
ML Alignment & Theory Scholars Program (MATS 9.0)
Simplex AI Safety, Berkeley
Interpreting Chain of Thought in Toy Models using Computational Mechanics. Mentors: Adam Shai & Paul Riechers
- 2025.07.01
Mentorship for Alignment Research Students (MARS 3.0)
Geodesic Research, Cambridge
Architectures for Increased Externalisation of Reasoning. Mentors: Puria Radmard, Cameron Tice & Edward Young
- 2025.05.01
Workshop: Interpretability in LLMs using Geometric and Statistical methods
University of Amsterdam
Organized with Jan Pieter van der Schaar
Awards
- 2020.09.01
Sander Bais Prize for Academic Merit
Institute for Theoretical Physics Amsterdam
For exceptional academic performance in the master's program
- 2015.05.20
ACM ICPC World Finals (Honorable Mention)
ACM
Represented India in the international collegiate programming contest
- 2013.05.01
International Olympiad in Informatics Training Camp
IARCS
Selected after ranking in top 25 in the Indian National Olympiad in Informatics
Skills
| Programming | |
| PyTorch | |
| TensorFlow | |
| Hadoop | |
| MapReduce | |
| Mathematica |
| HPC Resources | |
| LUMI consortium | |
| Snellius supercomputer |
Projects
- 2020.09 - 2021.08
Exploring the Spectral Theory/Topological Strings Duality (Master's Thesis)
This thesis analyzes the ST/TS duality, linking topological strings on toric Calabi-Yau manifolds to quantum mechanical spectral theory.
- 2017.01 - 2017.04
Real Space Renormalization and Applications to Machine Learning (Bachelor's Thesis)
We explore the application of renormalization group (RG) theory to understand the success of Restricted Boltzmann Machines (RBMs).