Entrance Desk Data Concierge assisted GTC 2026 guests with navigation and occasion data. | Credit score: The Robotic Report
In simply over a 12 months, IntBot Inc. has gone from idea to full‑physique humanoids greeting 1000’s of visitors at NVIDIA’s GTC and in lodge lobbies. The Sunnyvale, Calif.-based startup has used 24/7 interplay footage and sentiment evaluation to coach a social intelligence engine that sits on prime of off‑the‑shelf {hardware}.
At GTC 2026, CEO Lei Yang introduced that the firm‘s IntEng “normal social intelligence engine” now helps a number of humanoid and service robots from totally different {hardware} distributors. He stated this marked a big step towards hardware-agnostic deployment of socially clever robots in real-world environments.
This full-length view of the IntBot humanoid reveals its ft, though the robotic was stationary and secured to face for its shift on the GTC26 assist desk. | Credit score: The Robotic Report
IntBot additionally showcased the primary edge deployment of the NVIDIA Cosmos Cause-2 vision-language mannequin (VLM) inside its software program stack. Working straight on robotic edge compute methods, the mannequin allows robots to carry out real-time scene understanding, permitting them to interpret advanced human environments corresponding to crowded convention areas.
“The primary-generation robotic was a pre-programmed sort of motion. However for our robotic, in case you have been at CES or watched the movies, all of the feelings are generated,” acknowledged Yang. “So even in case you are not speaking to the robotic, the robotic would reply with some very pure, very refined movement, simply the very facet nodding, to indicate ‘Okay, I’m listening,’ and even the sort of movement to point ‘I’m alive.’ Every little thing is pushed by our social intelligence.”
Working straight on robotic edge compute methods, the mannequin allows robots to carry out real-time scene understanding, permitting them to interpret advanced human environments corresponding to crowded convention areas.
In line with Yang, ItBot’s robots use a type of audio-visual fusion, combining what it hears with what it sees, to higher perceive who within the scene is speaking and what the audio system’ intent could be. This permits the robotic to supply a extra pure interplay with people.
IntBot stays platform-neutral
Whereas most humanoid startups chase ever-better locomotion and manipulation, IntBot is intentionally staying “{hardware} agnostic,” positioning its software program as a social intelligence layer that may experience on prime of no matter platforms the business produces subsequent.
At present, that stack powers Nilo, a full-body humanoid that works 24/7 as a multilingual concierge in lodge lobbies from New York to Las Vegas, mixing on-device notion and body-language era with cloud LLMs for deeper queries.
“Proper now, we have already got three resorts throughout the U.S.,” stated Yang. “[We’re at] The Nap York in New York Metropolis, and a second one known as Otonomous in Las Vegas, and the third one is a Marriott Lodge in Tulsa, Okla. And all of those three robots function 24/7, mainly. They work alongside their human employees members, however IntBot provides add-on capabilities to enhance what the human employees, concierge, or [other] folks can do.”
By focusing first on noisy, real-world environments like CES and busy lodge lobbies, the place earlier kiosk-style methods and robots like Pepper stumbled, Lei Yang is betting that mastering pure, multi-party interplay would be the key to getting humanoids accepted as on a regular basis co-workers, not simply trade-show spectacles.
IntEngine coordinates notion, communication in actual time
Nylo’s potential to function autonomously on the GTC present ground is powered by IntEngine, IntBot’s proprietary, multimodal, multi-loop social intelligence system. IntEngine fuses imaginative and prescient, audio, and language in actual time to coordinate speech, facial features, and gesture—enabling robots to understand social context and reply naturally.
This structure permits Nylo not simply to reply, however to resolve when and the right way to have interaction, which Yang stated is a necessary functionality for working in open, public environments.
