I am an IMPRS-IS, Amazon Science Hub, and ELLIS PhD Fellow at the Deep Models and Optimization group, Max Planck Institute for Intelligent Systems and ELLIS Institute, Tübingen, supervised by Antonio Orvieto and Volkan Cevher (LIONS Lab, EPFL).
My research focuses on post-training methods and graph-based learning frameworks, with parallel interests in novel architectures — particularly hypernetworks and mixture-of-experts systems. Broader themes include code generation evaluation, reasoning, preference modelling, and mechanism design.
dheu.ai ଢ଼େଉ pron. "dheh-oo" [ɖʱeu]
Dheu (ଢ଼େଉ) means wave in Odia. I was born in Odisha, and across all my ventures I have encountered the synonymous symbolization of waves — from mountain climbing to research. This venture symbolizes that for me.
dheu.ai is an AI platform for Mergers & Acquisitions that combines game theory, graph intelligence, and multi-agent systems to model complex negotiation processes — providing advisory, risk mitigation, and strategy optimization.
Reach out at hello@dheu.ai
Before my PhD, I completed an MSc under the supervision of Irina Rish at Mila, Université de Montréal. I also founded Landskape AI (2019–2024), a research group on analytical deep learning theory, and spent time as a Machine Learning Engineer at Weights & Biases, a researcher at Morgan Stanley (continual learning, time series), and a visiting scholar at CMU's Human Sensing Lab and VITA, UT Austin. This June I am joining Aleph Alpha Research in Heidelberg as a research scientist intern.
I climb mountains — summited Jbel Toubkal (4167m, Morocco), Pico de Orizaba (5636m, Mexico), and Acotango (6052m, Bolivia). I've traveled to 62 countries across 6 continents, play contemporary and classical piano (1, 2), paint Madhubani and watercolors, and ran the Tübingen 100 km Staffellauf 2025 (rank 14).