Welcome
to the repo
I am an MSc student in Systems and Control at Delft University of Technology (TU Delft). My interests sit where control theory and nonlinear dynamics meet machine learning — and, increasingly, where those ideas extend from a single system to many independent agents that must coordinate without a central authority.
Previously, I was an AI Research Consultant at Nasiko AI, building open-source infrastructure — an agent registry and control plane — for discovering, governing, and coordinating autonomous AI agents. Before that I was a Project Associate in the Visual Computing Lab at the Indian Institute of Science (IISc), where I worked on federated learning across heterogeneous, unlabeled clients and on mechanism design for eliciting contributions from self-interested agents. I received my B.E. (Hons.) in Electronics & Communication from BITS Pilani, Hyderabad, where I also researched sparse nonlinear system identification for robots.
Links: Google Scholar · LinkedIn · GitHub
Research interests
- Control theory
- Nonlinear dynamics
- System identification
- Federated & decentralized learning
- Mechanism design
- Multi-agent systems
- Meta-learning
- Nonlinear optimization
Experience
- Built an open-source framework and control plane — an intelligent agent registry helping enterprises discover, govern, and coordinate autonomous AI agents.
- Designed the memory architecture for multi-agent flows to mitigate context blow-up.
- Designed persuasion mechanisms to elicit contributions from rational, self-interested agents in collaborative decentralized learning.
- Developed federated-learning algorithms to close domain gaps when fine-tuning pre-trained models on heterogeneous clients in unsupervised settings.
- Developed a sparse nonlinear system-identification algorithm for robots, accounting for changes in dynamics under varying environmental factors.
- Used model-agnostic meta-learning (MAML) to generalize to unseen dynamics.
- Reviewed robotics courses spanning the ROS2 Control framework, the Nav2 navigation stack, behavior trees, and MoveIt for manipulation.
- Edited the machine-learning course to teach concepts through ROS simulations.
- Built state-space models of a double-sided linear induction motor (DSLIM).
- Designed a DSLIM velocity controller using recurrent wavelet neural networks.
- Led development of the avionics subsystem and flight software; interfaced sensors and communication modules.
- Designed a gimbal controller to stabilize the probe's camera angle.
Publications
Education & awards
Contact and more
Last updated: July 2026.