Ramin Hasani is the co-founder and CEO of Liquid AI and a machine-learning scientist associated with MIT CSAIL. His research studies neural networks that model changing systems over time, including Liquid Time-constant Networks.[1][3]
Start with Liquid Time-constant Networks →
Education
Links
Liquid AI biography ORCID Website LinkedInCo-founder and chief executive officer · Liquid AI
Machine-learning scientist, CSAIL · MIT CSAIL
Career
Hasani completed his computer-science doctorate at TU Wien in May 2020. His subsequent work included postdoctoral research with Daniela Rus at MIT CSAIL and a principal AI and machine-learning scientist role at Vanguard, held alongside a CSAIL research affiliation. Liquid AI’s current team biography identifies him as its co-founder and CEO.[1]
Contributions
Liquid Time-constant Networks model a sequence as a continuously evolving hidden state. Their time constants change with the state and input through learned gates, and a differential-equation solver computes the output. This gives the model a way to adjust its response to changing temporal conditions. The paper combines mathematical bounds with time-series prediction experiments; those results concern the studied systems and do not by themselves establish general intelligence.[3]
Tools, writing & talks
Writing, research & talks
Start here
Liquid Time-constant Networksarxiv.orgA recurrent model whose time-dependent dynamics adapt through learned gates.
Begin with the model diagram and time-series experiments. Understanding recurrent neural networks and differential equations helps; use the linked code to inspect one example before transferring the method to a new control problem.
Co-author[3]