Parkinson's disease (PD) affects over 10 million individuals worldwide, with motor symptoms including tremor, bradykinesia, and freezing of gait (FoG) significantly impacting quality of life. While commercial assistive robotics exist, costs ranging from $3,000 to $70,000 create prohibitive barriers for most patients and caregivers. This paper presents ARIA (Adaptive Robotic Intelligence for Aging), an open-source framework for building accessible AI-powered companion systems for individuals with Parkinson's disease. We synthesize current research in soft robotics, socially assistive robots (SARs), and conversational AI to propose a family of six companion devices—spanning tabletop companions to walker-integrated mobility assistants—all buildable for under $500 using commodity hardware. Drawing on evidence that external visual and auditory cueing can reduce freezing of gait episodes by up to 68%, we detail the neuroscientific basis for our intervention mechanisms and present a democratized approach to assistive technology development. Our analysis of the $52 billion annual PD care market reveals significant opportunity for open-source disruption, with particular impact potential in underserved populations lacking access to specialized care.