India’s physical AI race has a new entrant. Inner Sky Labs, the company behind Miko, has unveiled India-built foundation models meant to help machines see, hear, feel, decide and act. The October 2026 stack includes Drishti for vision, Shruti for audio, Sparsh for touch, action models Kriya, Prana and Karma, and a safety layer called Kavach.
The timing is notable. India installed almost 10,500 industrial robots in 2025, up 15%, according to the International Federation of Robotics. India remained sixth globally, while installations grew at an average 27% a year between 2020 and 2025.
Key Highlights
- Drishti handles vision, Shruti audio and Sparsh tactile input.
- Kriya, Prana and Karma turn perception into planned machine actions.
- Kavach checks actions against safety rules before execution.
- Inner Sky Labs is closing a ₹300 crore funding round for data, compute and training.
What Is Physical AI And How Is It Different From Generative AI?
Physical AI is artificial intelligence built to operate through machines in physical environments. A chatbot produces text or images. A physical AI system may have to spot a worker beside a robotic arm, interpret a spoken instruction, choose a safe movement and control hardware.
The Office of the Principal Scientific Adviser’s August 2026 Physical AI edition describes the shift as AI moving beyond screens into robots, drones and autonomous systems that can sense surroundings, reason, plan and act. The office also discussed India’s Physical AI push through its official social update.
That raises the stakes. A bad chatbot response stays on a screen; a bad machine decision can move equipment incorrectly or put people and objects at risk.
How Do Drishti, Shruti And Sparsh Give Machines Perception?
Inner Sky Labs calls Drishti, Shruti and Sparsh its perception layer. Each handles a different stream before the machine decides what to do.
- Drishti: Processes visual scenes and objects from cameras.
- Shruti: Interprets sound, including spoken instructions and surrounding audio.
- Sparsh: Uses tactile information so machines can react to touch, force and contact.
Sight alone may not be enough for tasks such as gripping. A robot can identify an object correctly but still need tactile feedback to decide how firmly to hold it.
The company says these models draw on nearly a decade of physical-world data from factories, Miko’s manufacturing ecosystem and other environments. CEO Sneh Vaswani told Inc42’s report on the Inner Sky Labs launch that they were developed in-house rather than fine-tuned from an open-source foundation model.
What Happens After A Machine Sees, Hears Or Feels Something?
Perception is only the start. Kriya, Prana and Karma are designed to help plan movements and predict what could happen next. Inner Sky Labs says its first generation is already used in industrial applications combining vision, robot position, movement, force and torque data.
Kavach then evaluates a proposed action before execution. The company says the safety layer has more than 20 guardrails, including checks for human presence, movement speed and permitted operating areas.
A factory sequence could look like this:
- Drishti identifies a component and nearby space.
- Sparsh helps interpret contact and force.
- Kriya, Prana or Karma plan movement.
- Kavach can block the move if a safety rule fails.
The platform is described as hardware-agnostic and available through cloud APIs, on-premise systems or on-device processing, including offline environments. That approach could appeal to industrial or government users that want operational data kept locally.
Why Does This Launch Matter For India’s Physical AI Push?
Inner Sky Labs was founded in 2015 by IIT Bombay alumni Sneh Vaswani, Prashant Iyengar and Chintan Raikar. Miko became its public-facing product while, according to the company, the broader physical AI effort developed over roughly a decade.
The startup is now closing a ₹300 crore round. Inc42 reported that completion would take total funding to about $100 million, with fresh capital planned for physical-world data collection, compute and model training.
India’s robotics base is expanding at the same time. The Principal Scientific Adviser’s 2026 work on Physical AI has highlighted domestic activity across industrial automation, drones, underwater systems, sanitation and humanoids. The harder test comes after launch: whether locally built models deliver reliable performance and independently verified safety across changing environments.
What is Physical AI?
Physical AI combines artificial intelligence with robots or machines that sense surroundings, make decisions and perform actions through hardware while operating inside changing physical environments.
What is the Drishti AI model?
Drishti is Inner Sky Labs’ vision-model series, built to process camera input and help machines identify scenes, objects and visual conditions before planning physical actions.
What does the Shruti model do?
Shruti is the company’s audio-perception series, processing speech and surrounding sounds so a machine can use audio cues while deciding how it should respond physically.
What is Sparsh used for in Physical AI?
Sparsh processes tactile information, helping machines react to touch, pressure, force or contact when a physical task needs more feedback than cameras alone can provide.
Is Inner Sky Labs connected to Miko?
Yes. Inner Sky Labs is the umbrella brand behind the Miko business, and its founders say years of work on Miko helped develop the wider physical-AI stack.
AI Breakthroughs and Tech Discoveries
1. What Are AI’s Hidden Environmental Costs?
Uncover how AI growth affects electricity, water and land resources.
2. Can IISc’s Challenge Spark New Innovations?
Explore the IISc Seva First Innovation Challenge 2026 and its focus on new ideas.
3. How Does GalaxEye’s Dual-Sensor Imaging Work?
Discover GalaxEye’s satellite imaging technology and its US patent.
4. What Makes Gemini 3.7 Flash Different?
Check out Google’s Gemini 3.7 Flash features and AI capabilities.
5. How Power-Hungry Are AI Data Centers?
Find out how AI data centers consume electricity and affect energy demand.
