Sensing + grounding
Fuse vision, audio, mmWave, environmental sensors, and device events into evidence tied to real rooms, people, and objects.
AI-native home
Build the brainfor the whole home.
LiveLarge is building a persistent intelligence layer for the physical home—one that understands residents and space, remembers change, coordinates devices, and enables robots to act with context and permission.
The system
The house is not a collection of smart devices.
One home. One ambient system.
Today, every device sees only its own sensors, and every robot enters a home almost blind. HomeOS creates the missing shared layer: a complete semantic spatial Home Model, long-term Home Memory, identity and permission, planning, execution, and verification.
HomeOS architecture
A home brain grounded in reality.
What the home brain must contain
Home Model
Maintain a continuously calibrated semantic spatial model of the entire home—not isolated device states or a single camera view.
Home Memory
Remember identity, state, routines, changes, and confidence over time, with retrieval, correction, replay, and evaluation.
HomeOS Core
Turn intent into permission-aware plans, coordinate execution, verify physical outcomes, recover from failure, and write results back.
Assist + interfaces
Give people, devices, services, and robots one understandable way to ask, clarify, act, report, and remain under user control.
How HomeOS closes the loop
Physical homeRooms · people · objects · sensors
Lila SpaceAIComplete Home Model + Home MemoryGrounding · identity · state · time · evidence
Lila Core · HomeOS brainPlan · authorize · execute · verify · recoverHuman intent remains connected to physical outcomes
Action layerDevices · services · robots
Continuous learning loopOutcome evidence updates the Home Model + Home Memory
Closed-loop household intelligence means every action is grounded, authorized, executed, verified, and written back—so the home becomes safer and more capable with use.
Home × robot
A robot brings the body. HomeOS gives it the home.
A capable humanoid robot brings locomotion, manipulation, perception, and its own local planner. But the body alone still does not know this home's people, object history, device state, privacy rules, or what ‘done’ means to the resident.
HomeOS translates a resident's intent into a synchronized task contract: who and what are involved, where to act, what is allowed, which devices and services are available, and how completion will be verified.
The robot can focus on physical execution while HomeOS maintains household context, authority, coordination, and memory.The home is the ultimate environment for embodied intelligence.
The home is not a backdrop. HomeOS gives embodied agents the spatial model, memory, permissions, device state, and success criteria they need to work safely across real life.
Who + what
Person, object, and device IDs; current location, state, movement, affordances, confidence, and evidence.
Where
Map version, coordinate frame, room topology, target pose or region, reachable areas, and dynamic obstacles.
State + memory
Relevant Home Memory, recent changes, and the live state of household devices and services.
Permission + safety
Authorization scope, privacy boundaries, forbidden zones and actions, confirmation requirements, and safety policy.
Task + success
Task ID, structured goal, constraints, dependencies, validity, completion conditions, fallback, and recovery.
Return channel
Accepted, running, paused, completed, or failed; pose, progress, evidence, exceptions, verified outcome, and write-back.
Join LiveLarge
Help the home
understand reality.
Tell us what you have built, which part of the system you want to own, and why intelligence should leave the screen.
TeamCore engineering teams in Guangzhou and Beijing
ScopeSeniority and role shaped around system ownership
ProcessWe reply with team, location, and interview details when there is a fit