Karthik Viswanathan

Doctoral researcher at Helmholtz Munich and TUM, in the group of Fabian Theis.
My interests are genomics, representation learning, flow matching, and single-cell perturbation prediction.
Papers are listed below and on Google Scholar.
Contact: karthik.viswanathan@helmholtz-munich.de, or krthkvswnthn@gmail.com.

Papers and preprints

scConcept: Contrastive pretraining for technology-agnostic single-cell representations beyond reconstruction.
M. Bahrami, A. Tejada-Lapuerta, S. Becker, M. Shulman, K. Viswanathan, D. Sharma, F. S. Hashemi G., F. J. Theis.
Nature Methods, under revision.
BioToken and BioFM: Biologically-informed tokenization enables accurate and efficient genomic foundation models.
A. Medvedev, K. Viswanathan*, P. Kanithi, K. Vishniakov, P. Munjal, C. Christophe, M. A. F. Pimentel, R. Rajan, S. Khan.
ICML, 2026. *Co-first author.
Tokenization to transfer: Do genomic foundation models learn good representations?
K. Vishniakov, K. Viswanathan, A. Medvedev, P. Kanithi, M. Pimentel, R. Rajan, S. Khan.
ICLR, 2026.
Gene42: Long-range genomic foundation model with dense attention.
K. Vishniakov, B. Ben Amor, E. Tekin, N. A. ElNaker, K. Viswanathan, A. Medvedev, et al.
NeurIPS Workshop on AI for Virtual Cells and Instruments, 2025.
Ecological resilience in a circular world: Mutation and extinction in five-species ecosystems.
K. Viswanathan, A. Wilson, S. Bhattacharyya, C. Hens.
Chaos, Solitons & Fractals 180, 2024. First author.
Streamlining pipeline efficiency: A novel model-agnostic technique for accelerating conditional generative and virtual screening pipelines.
K. Viswanathan, M. Goel, S. Laghuvarapu, G. Varma, U. D. Priyakumar.
Scientific Reports 13, 21069, 2023. First author.

Decomposing a raster image into constituent elements utilizing discrete layering and classification.
A. Agarwal, J. K. Jose, K. Viswanathan, B. V. Srinivasan, D. S. Shah, M. R. Wayal.
Adobe Inc. US patent application US 2026/0051064 A1, filed 2024, published 2026.