Medical Imaging · Robotic Perception · Endovascular Autonomy

Liam McHugh

I'm an algorithms engineer focused on medical imaging, robotic perception, and endovascular autonomy. My experience spans vascular feature extraction, anatomy-aware state estimation, physics & sensors simulation, ML/RL workflows, & C++/Python systems for cyber-physical medical devices.

I'm also an avid cyclist, climber & hiker. Feel free to reach out with advice, ideas, or just to get outside!

View projects Portfolio PDF GitHub LinkedIn

GraduateMS Electrical Engineering, Columbia
UndergraduateBS Mechanical Engineering, UC Berkeley
Based inSan Francisco, CA

Experience

Mosaic Intelligence Labs Inc.

Founding Engineer / Member of Technical Staff

Dec 2024 – Present San Francisco, CA

Building prototype algorithms for image-guided endovascular robotics — vascular perception, physics/interaction modeling & partially-observable state estimation — including probabilistic factor-graph methods for joint instrument-vessel localization & mapping under deformable anatomy & uncertain sensor/robot observations.
Prototyped RL/behavior-cloning workflows for vascular navigation using simulated anatomy, expert trajectories, & safety-oriented evaluation metrics.
Built Python/C++/ROS2 tooling for experiment orchestration, benchtop system integration, & medical data pipelines.
Translated clinician feedback from OR walkthroughs & physician interviews into UI/UX prototypes, workflow requirements, & product/engineering direction.

Technical Leadership

Liquid-powered rocket on the launch rail

Project Manager: Liquid-Powered Rocket

UC Berkeley Space Technologies & Rocketry

Led a 30+ member student team to develop launch control system and liquid-propulsion flight vehicle.

Learn more

Cardiac flow diagram used in a fluid dynamics recitation

Chemical vapor deposition reactor schematic

Teaching Assistant: Fluid Dynamics

UC Berkeley

Weekly recitations applying fluid dynamics principles to real-world applications.

Teaching Assistant: Graduate Mechatronics

Columbia University

Lab setup & assistance for advanced mechatronics & microcomputer systems.

Academic Research

Surgical robot manipulators in a training rig

AI & Deep Learning Lab, Columbia

Jan 2025 – present

Surgical Robotics: Behavioral Kinematics Assessment

Advisor: Prof. Zoran Kostic & Dr. Filippo Filicori, Northwell Health

Kinematics-driven intelligence for tele-operated & autonomous surgical robotics behavioral anomaly detection.
Instructive foundation model: Pretraining on auto-annotated operations with finetuning on instructive semantics.
Latency cost optimization for adaptive edge task offloading. Information & communication theory fundamentals.

Battery state of energy plotted against locational marginal price

Distributed Energy Resource Management

Goldman School of Public Policy / General Motors

Vehicle-to-grid battery arbitrage strategies using sota ML techniques. Guided deployment policies for commercial & utility EV sectors.

Renewable Energy Resources robust portfolio management for large independent consumer load sites.

Laser Thermal Lab research overview figure

Berkeley Laser Thermal Lab

Model Diagnostics of Phase Explosion in Femtosecond Laser Ablation

Ablation event energy transport model & supercontinuum generation methods for pump-probe scattering spectroscopy. Ejecta morphology control w/applications in micromachining, energy catalysis and nanoscale medical treatments.

Engineering Projects

MUSE service architecture: wearable edge device feeding a music stream processor

MUSE: User-Sensitive Musical Expression

Independent research, lead developer · New York, NY

Activity-Sensitive Generative Music: Fusing physical activity data with local generative audio models for controllable, realtime activity-music generation.

Learn more

Neural Guitar Pedal

Wireless headband EMG for neural interface to guitar pedal effects modulation. Embedded STM32 signal acquisition system w/BLE comms to main pedal unit for real-time ICA & ML-based feature extraction.

Learn more