We show how illiquid prediction markets can be made more feasible by learning a joint distribution across multiple markets. We show why parlay-style markets are a crucial building block to achieve that.
We develop an axiomatic theory for AMMs in local energy sharing markets and analyze the Markov Perfect Equilibrium of the resulting economy with a Mean-Field Game
We show how yield tokenization can mitigate interest rate risk for lending in DeFi, design AMMs for their price discovery, and design a fixed rate lending protocol
We explore the tradeoff between adaptivity and robustness in automated lending protocols
Adapting the interest rate and collateralization factor of a decentralized lending platform to ensure solvency and stability, using a two-timescale control algorithm
Using canonical control theory to derive optimally efficient adaptive bonding curves for automated market making
Using reinforcement learning to design optimal market making schemes for Decentralized Finance
Using curriculum learning to learn near-optimal decoders for PAC and Polar codes.
Senior Year Project - 2020 (IIT-Bombay)
On opinion shaping of populations in the multi-armed bandit framework
RoboSub 2018 (San Diego)
Technical Design report of the Matsya 5 vehicle I worked on while at IIT-B
Summer 2017 (IIT-Bombay)
Fun summer project on using non-linear dynamics to model the syncing of fireflies, the spread of internet memes and love triangles