Arnav Adhikari

I am a student at MIT studying electrical engineering, computer science, physics, and mathematics. I currently do robotics research with the Improbable AI Lab, where I am advised by Nolan Fey and Pulkit Agrawal. My research focuses on agile loco-manipulation through sim-to-real RL and whole-body control. I'm currently exploring long-context adaptation and latent inference for handling dynamics mismatch at deployment. I'm grateful to be supported by the Paul E. Gray UROP Fund.

In the past, I've done research in cardiac imaging with leading medical institutions (Baylor College of Medicine / Texas Children's Hospital / Stanford AIMI), AI surgical robotics (Rice / Harvard Medical School), and robot motion planning. I've also built robots and algorithms for VEX World Championships, and am a retired USAJMO winner and USACO Gold Medalist.

Email / Scholar / Github / X / LinkedIn

Selected Research

I am interested in embodied intelligence, robotics, nanotech, photonics, edge computing, neurotech, and synthetic biology.

	Building Production-Grade Evals for Multi-Agent Systems Kyle Tianshi, Arnav Adhikari, Anna Monaco, June Lee Paradigm AI We built and scaled a production-grade evaluation system for a multi-agent platform operating thousands of concurrent agents across 30+ use cases, positioning the product as a replacement for traditional SDR, GTM, and PM workflows. Engineered an Agent-as-a-Judge system tracing multi-step reasoning chains — verifying step validity, source relevance, factual correctness, and solution-path optimality.
	Automated Assessment of Left Ventricular Function in Tetralogy of Fallot - A Machine Learning Approach Arnav Adhikari, Giles Vick, Minh B. Nguyen, Tam Doan, Mounica Y. Rao, David Ouyang, Anitha Parthiban, and Lalita Wadhwa WCPCCS 2023, Circulation AHA Research on interpretable machine learning for cardiac imaging, aimed at making model predictions more transparent and clinically useful.
	Towards a Web-Based Digital Twin for the Cardiac Operating Room Arnav Adhikari, Sangwon Seo, Marco Zenati, Roger Dias, Eduardo Salas, Vaibhav Unhelkar Rice AIHC 2023 Analyzing team dynamics and synergy to inform AI-driven coaching for surgical robotics.
	Robot Motion Planning & Control Arnav Adhikari, Kaiyu Hang Work on motion planning and control for robotic systems.
	harmoNIC (HackMIT 2024 Winner) Arnav Adhikari, Arjun Melwani, Tahsin Zaman HackMIT, 2024 A smart nicotine patch built at HackMIT 2024.
Venture Cola Coaches	Cola Coaches Arnav Adhikari Founder colacoaches.arnavadhikari.com A venture I founded. Also run Pro Tutoring services.

Coursework

6.7800 (G): Inference & Information
6.2080: Semiconductor Electronic Circuits
6.4210: Robotic Manipulation
6.191: Computation Structures
6.4110: Representation, Inference, and Reasoning in AI
18.650: Fundamentals of Statistics
18.600: Random Variables and Probability

Design based on Jon Barron's website.