01 — Active SLAM — Autonomous Map Exploration Active SLAM autonomous mapping agent
02 — Robot Manipulation — Learning from Demonstration Robot learning from demonstration

Tvisha Shah

CS / CE · Brown University · tvisha_shah@brown.edu · Resume ↗

I love decision making — currently exploring that through intelligent agents and robots using policies. I'm also interested in cognitive science, and am currently a research assistant at the Carney Institute of Brain Sciences at Brown. I love working at the intersection of minds and machines.

Tvisha Shah

my work

01

Autonomous City Mapping with Multi-Agent Active SLAM

Three RL agents with Mamba state-space memory collaboratively build a real-time city map from raw sensor data alone, achieving 90%+ coverage on real-world street grids.

02

Hindsight-Driven Self-Improvement for Robot Manipulation

Iterative policy improvement through hindsight-relabelled self-rollouts on goal-conditioned manipulation tasks.

03

Zero-Shot 3D Affordance Mapping

Language-driven affordance localization in 3D point clouds. CLIP × geometry priors — no training data required.

04

Latent Stress Dynamics in Sleep

Computational study of HRV-derived stress states and sleep architecture across 8 months of longitudinal wearable data. Accepted at CCN 2026.

I read, play 4 instruments, write poetry, do debates and public speaking, run an NGO in India, and go to the gym.  Looking to work with a small and passionate team!