Research Initiative
Embodied CARLA: a robotic care assistant
We're researching how CARLA — the AI care agent behind InclusiCare — can power an in-home robotic assistant that observes and documents the care routine on-device, relieving caregivers of the documentation burden and assembling what it sees into a care repository parents and clinicians can actually use.
The problem: caregivers are the sensor and the scribe
Families of children with developmental disabilities carry a double load: they provide the care, and they document it. Behavioral incidents, triggers, sleep, meals, medication responses, what worked and what didn't — the record that clinicians, schools, and future caregivers depend on only exists if an exhausted parent writes it down. Most of it is never captured, and what is captured is fragmented across notebooks, texts, and memory.
CARLA already reduces that burden in InclusiCare by turning natural conversation into structured care records. Our research asks the next question: what if the documentation didn't depend on the caregiver at all? An embodied assistant in the home can observe the care routine as it happens, capture the patterns caregivers are too busy caring to record, and give families back the hours they spend reconstructing their day.
Research pillars
Observation, not intervention
The robot assists the caregiver — it does not deliver therapy or make clinical decisions. It watches the routine, documents what matters, and stays out of the way. Care remains entirely in human hands.
CARLA as the understanding layer
The same extraction-first, safety-hardened AI agent that powers InclusiCare interprets what the robot observes: distinguishing a behavioral incident from play, recognizing each child's patterns, and structuring observations into the family's care record.
The care repository
Observations flow into a longitudinal, FHIR-structured care bank the family owns. Over time it becomes the shared evidence base that lets parents spot patterns, clinicians see between appointments, and every new caregiver start informed.
Privacy as architecture
Observation happens in the most sensitive place there is — a child's home. Perception and inference run on-device; raw audio and video never leave the house. Families control what is recorded, what is kept, and who sees it.
Technical foundation
This research is designed around the NVIDIA robotics and simulation stack. InclusiGear is a member of the NVIDIA Inception program, which supports this work.
Jetson edge compute
On-device perception and inference are the foundation of the privacy model: behavioral data is processed locally, and only structured, family-approved observations reach the care repository.
NVIDIA Isaac + Omniverse
Home-environment observation scenarios are developed and stress-tested in simulation first — evaluating safety, accuracy, and unobtrusiveness before hardware enters any family's space.
CARLA safety pipeline
Our existing evaluation and safety-probe framework for CARLA extends to embodied observation: bias probes, drift gates, and scenario-based red-teaming of what the system records and reports.
Grounded in evidence
Published research on robots in the lives of children with developmental disabilities — including large public datasets from clinical robot-assisted intervention studies — shows that children engage comfortably with robotic companions in structured settings. We build on that evidence for acceptance and safety, and on our own caregiver discovery research documenting where the documentation burden falls hardest. The evidence base behind InclusiCare — caregiver-facing language models, clinical speech structuring, behavioral pattern detection — is the same foundation this work extends into the physical world.
Where this stands
This is an early-stage research initiative, not a product — and deliberately not a medical device. It is a care-coordination and documentation tool, the same hard line we hold across the InclusiGear platform: AI structures information; humans make every care decision. Our current focus is simulation-first feasibility work, identifying clinical and academic collaborators, and pursuing grant funding to support the research phase.
Research milestones and findings will be published here as the work progresses.
Interested in this work?
We're speaking with investors, clinical partners, and researchers interested in robotic caregiving support for the developmental disability community.
Get in touch