Holograms + conversational AI
AI that shows up.
Digital AI avatars that look human, answer clearly, respond naturally and help someone move forward.
Book a demo
Original R&D
From DECS to deployed digital people
Granted patent
Indian Patent No. 579493
In practice
Darwinbot and NARI, deployed
DECS
AI in the form of a person they can talk to.
DECS powers the conversation. Holobot gives it a place, as a hologram, in the room. Built around a consenting person or approved character, it draws on knowledge you approve and works within boundaries you set.
Their face
Captured with consent, so people see and hear an identity they recognise.
What they may say
Grounded in approved sources, with clear boundaries and a handoff when a question goes beyond them.
How they answer
Voice, expression and timing adapt to the meaning of the reply.
One presence. Three ways to show up.
On screen, on a counter or at full height.
Digital
On screen
In a browser or app, or on an existing kiosk screen.
No dedicated device to ship.
Compact
Miniature Holobot
A holographic unit for counters and one-to-one demos.
Where space is tight.
Full height
Life-size Holobot
Full height interactive hologram for lobbies, stands and launches.
People walk up to it.
Consented face + voiceUltra-low latencyMultilingualAPI-connected
Proven in the world
One system. Two very different jobs.
Darwinbot guides product choices.
Ask Darwin. Watch the work happen.
Visitors asked in their own words. Darwin opened the live Darwinbox product and walked them through the right workflow.
View project →
NARI puts expertise within reach.
One surgeon. Many languages.
NARI let people ask breast-health questions in their own language, extending Dr. Raghu Ram’s guidance far beyond the room.
View project →Then we pushed it further with response-aware expression.
A looped talking head performs every answer the same way. DECS doesn’t.
Recreating a person is only the start. Each answer shapes the voice, expression, gaze and timing; so reassurance feels calm, warnings land clearly and good news feels like good news.
Meaning-aware voice, face and timing—in 1080p, multilingual and API-connected.
How we build it
First the job. Then the face.
We define the knowledge, boundaries and outcome before we capture anyone. Then we test the hard cases and deploy in the form the job needs.
01 Define the job
Set the job, approved sources, limits and human handoff.
02 Capture with consent
Film the person and train the face and voice models.
03 Test the hard cases
Check answers, expression, languages, latency and failure handling.
04 Deploy and improve
Launch on screen or in a Holobot; refine with real use.
Where it can work
When people need to feel engaged — not just informed.
Product guidance
Let people ask what they need, then walk them through the right path in a product.
Expert access
Make approved specialist guidance available across languages and places.
Service and training
Give staff or visitors a clear explanation they can question, not just watch.
Indian Patent No. 579493
Granted 30 January 2026
System and method for producing an audio-video response to an input.