As industries embrace AI-led automation, the physical world inside factories remains largely invisible. Kaushal Kukkar of AmiMotion tells EFY’s Akanksha Sondhi Gaur how the company is making factories searchable with a full-stack Physical AI platform that combines UWB sensing, real-time location intelligence, and AI-driven analytics.
Q. What inspired the creation of AmiMotion, and what gap did you identify?
A. We identified two fundamental problems across industrial environments. First, workers in factories and warehouses often spend more than an hour each day searching for trolleys, tools, scanners and other assets, resulting in roughly 15% productivity loss.
The second is worker safety. In automotive plants, oil and gas facilities and other industrial sites, organisations often do not know the exact location of workers. During emergencies, evacuation can take more than 30 minutes, making timely response difficult.
Outside buildings, navigation systems make locations instantly visible. Inside large industrial facilities, that visibility disappears. We saw an opportunity to solve both operational inefficiency and safety challenges through ‘physical intelligence’.
Q. Why is AmiMotion a ‘Google for the physical world’?
A. We are building a search engine for the physical world; what Google did for the internet, we are now doing for factories by making the invisible inside them visible. AmiMotion is building a ‘search engine for the physical world’, much like Google made digital information searchable. Factories, warehouses, airports and industrial plants contain billions of moving entities, including machines, forklifts, tools, raw materials, products and people. Yet there is no system that makes them searchable or visible in real time.
Our platform continuously tracks assets, workers and machines, creating structured, real-time physical-world data that digitises movement and interactions. This transforms industrial environments into searchable, intelligent spaces. Our long-term vision extends beyond tracking. We are building the foundational intelligence layer for Physical AI systems that will power the next generation of autonomous, connected industrial operations.
Q. Why focus on Physical AI rather than conventional AI?
A. Traditional real-time location systems (RTLS) answer one question: ‘Where is something?’ Ambient Intelligence goes much further by explaining what is happening, why it is happening, what is likely to happen next, and what action should follow. We combine location data with workflow intelligence, Artificial Intelligence (AI), analytics and contextual understanding. Instead of simply showing a forklift at one location, the system detects congestion, predicts delays and recommends or initiates corrective actions.
Tracking becomes intelligence when location data evolves from coordinates into context, prediction and actionable insight. Unlike digital AI systems based on large language models (LLMs), our platform operates in the Physical AI domain, where intelligence depends on real-time, centimetre-level three-dimensional spatial data. We use spatio-temporal graph neural networks (STGNNs) to understand spatial relationships, movement patterns, human-machine interactions and operational behaviour in real time. This enables predictive intelligence and autonomous decision-making rather than conversational responses.
Q. When does tracking become intelligence, and how does the technology stack work?
A. The system begins with AmiGo ultra-wideband (UWB) tags attached to people and assets. AmiMo anchors installed throughout a facility receive their signals, while location is calculated using time-of-flight and multilateration techniques to generate real-time positional data. This flows into the AMI Hub Core, which creates a live digital map of operations. The application layer then delivers intelligence through two platforms.
AMI Fleet manages orchestration, including task allocation, fleet coordination and automated operations. AMI Sense provides analytics on worker movement, space utilisation, bottlenecks and process efficiency.
Overall, the platform integrates UWB, Bluetooth low energy (BLE), Wi-Fi and third-party systems into a unified end-to-end stack, from sensing hardware to AI-driven operational decision-making across industrial environments.
Q. What is your true competitive moat: hardware, data or the intelligence layer?
A. Our strongest intellectual property (IP) is not the hardware but the intelligence layer built on top of it. Hardware can eventually be replicated. What is much harder to replicate is the combination of proprietary algorithms, sensor-fusion models, positioning systems and AI-driven analytics that convert raw spatial data into operational intelligence.
Our competitive advantage comes from the quality of our centimetre-level, real-time three-dimensional movement data across complex industrial environments. We also combine information from robots, global positioning system (GPS), radio frequency identification (RFID), human activity and fleet systems into a unified framework, creating richer operational insight and stronger predictive capability.
Even if the hardware is copied, the accumulated data, deployment experience and intelligence layer create a durable competitive moat.
Q. Why choose UWB over other tracking technologies?
A. Every technology has its strengths. GPS performs poorly indoors, RFID provides zone-level visibility, and BLE offers a lower-cost but less accurate alternative. UWB delivers centimetre-level accuracy, low latency and strong resistance to interference, making it well suited to industrial applications where precise positioning drives operational decisions.
Customers typically move beyond BLE when they need more than simple proximity detection. While BLE can indicate whether an asset is nearby or within a general zone, it cannot reliably support precise positioning, movement analysis, object interaction or automation. As industrial workflows become more sophisticated, organisations increasingly adopt UWB because it provides decision-grade spatial intelligence rather than approximate location.
Q. How do you achieve 10cm accuracy, and how does the system perform in worst-case conditions?
A. Industrial environments are challenging because of metal surfaces, dense infrastructure and signal reflections. Accuracy is achieved not through hardware alone, but by combining UWB time-of-flight measurements with multilateration, filtering and correction algorithms to compensate for noise and non-line-of-sight conditions.
Real-world performance is affected by metal structures, electromagnetic interference, crowded layouts and constant movement. While laboratory conditions may demonstrate peak accuracy, the more important measure is consistent performance in live industrial environments. We achieve this through advanced calibration, sensor fusion and robust algorithms, ensuring dependable operation and actionable intelligence even under the most demanding conditions.
Q. What were the biggest design challenges, and how did you achieve such a small form factor with long battery life?
A. One of our biggest challenges was developing a system capable of tracking everything, from large industrial machines to very small handheld devices. This required extreme miniaturisation, resulting in what we believe is one of the smallest industrial-grade UWB tags, measuring just 3cm × 3cm.
Scalability presented another challenge, particularly in metal-rich factories, semiconductor facilities and dense industrial environments. We addressed this by developing a vendor-agnostic architecture that integrates multiple systems instead of relying on a closed ecosystem.
Battery life was equally critical. Rather than increasing battery capacity, we reduced power consumption through adaptive operating modes, minimising communication and computation when idle and activating higher-power states only when required. This delivers multi-year battery life without compromising responsiveness.
Q. Was the technology built in-house, and what role did academia play?
A. Yes. Everything was developed internally. I bring more than 20 years of experience in embedded systems, having worked at Intel and Texas Instruments, while my co-founder has around 25 years of experience in radio frequency and wireless hardware systems. I lead software and product development, while he leads hardware architecture. Together, we built the entire technology stack in-house rather than relying on imported designs.
Academia also plays an important role. Large industrial deployments, often spanning thousands of square metres, require rigorous validation before commercial rollout. Academic institutions provide testing infrastructure, simulation environments and expertise in algorithms and system design, helping us accelerate innovation while ensuring real-world reliability.
We already collaborate with the Indian Institutes of Technology (IITs) and are exploring deeper engagement with the Indian Institute of Science (IISc) to strengthen research and technology development.
Q. What fundamentally differentiates your platform from competitors?
A. We provide a complete end-to-end platform rather than standalone hardware, with a focus on Ambient Intelligence instead of simple location visibility. Our priority is rapid deployment and the delivery of measurable operational outcomes. Customers choose AmiMotion because they need more than tracking. They require accurate real-time visibility, seamless integration with existing operations and insights that improve efficiency and decision-making. Our platform enables them to move beyond knowing where assets are to understanding what is happening, why it matters and what action should be taken next.
Q. How do you scale from pilots to enterprise deployments?
A. Pilot deployments may involve a few hundred tags, while enterprise roll-outs typically require thousands across large facilities. At that scale, challenges include network synchronisation, infrastructure planning, latency management and maintaining consistent performance.
To address this, we developed a distributed architecture that scales efficiently across facilities of varying sizes and levels of complexity.
Unlike traditional proprietary systems, our platform is hardware-, data- and software-agnostic. It integrates with robots, smartphones, Global Positioning System (GPS), Radio Frequency Identification (RFID), existing industrial infrastructure and other data sources, standardising them into a unified intelligence layer.
This allows AI models trained in one environment to be reused in others with minimal reconfiguration, improving scalability, reducing deployment time and supporting rapid expansion across industries and applications.
Q. What did your first large deployment teach you?
A. Our first large-scale deployment confirmed that real-world environments behave very differently from controlled test conditions.
Although the core technology performed well, scaling introduced challenges such as signal reflections, environmental interference, infrastructure constraints and operational variability that affected consistency.
We learned that successful deployment depends not only on accurate hardware, but also on installation strategy, calibration, software intelligence and alignment with customer workflows. Those lessons led us to strengthen our algorithms, refine our deployment methodology and build a more resilient, scalable platform capable of performing reliably across diverse industrial environments.
Q. What insights deliver the biggest operational gains for customers?
A. In warehouses and manufacturing facilities, customers often discover bottlenecks they were previously unaware of. Heat maps reveal inefficient forklift routes, unnecessary worker movement and underutilised spaces.
The biggest gains come from replacing limited visibility with real-time, data-driven decision-making. Instead of relying on manual tracking or assumptions, customers gain immediate insight into assets, workflows and movement patterns. This reduces search time, improves resource utilisation, minimises inefficiencies and speeds up decision-making, transforming operations from reactive execution to proactive optimisation.
Q. How do you ensure scalable, high-quality manufacturing?
A. Our vision is closely aligned with building strong indigenous technology capabilities. While we source semiconductor components from companies such as NXP, Infineon and Texas Instruments, all core technology development is carried out in-house.
Critical elements, including printed circuit board (PCB) design, module design, enclosures and overall hardware architecture, are developed in India, with most engineering based in Bengaluru.
For manufacturing, we follow an Apple-like model, outsourcing production to specialised manufacturing partners. This allows us to focus on product design, innovation and core technology development while leveraging external expertise for efficient, large-scale manufacturing.
Q. How do you solve operational inefficiencies and scale globally?
A. Most warehouses still rely heavily on manual barcode or RFID scanning, requiring workers to stop, scan, confirm and update systems. We eliminate these manual steps by automatically detecting movement and recording location events in real time.
The broader challenge is that AI cannot operate effectively without high-quality physical-world data. We generate that foundational data layer and convert it into actionable operational intelligence.
Our business follows a hybrid product-led model that combines hardware and software-as-a-service (SaaS). We develop the core hardware, including UWB tags and anchors, while intelligence is delivered through subscription-based software. Deployments are handled by system integrators rather than a services-heavy internal model.
International expansion is partner-led. We already work with partners across Asia-Pacific, Europe and North America, and plan to continue expanding this network to support global growth.
Q. Where has early traction been strongest?
A. We are still at an early stage of our growth journey, with deployments already live at organisations including Maruti Suzuki, Volvo, Marmon and Tata Motors. Current deployments typically involve between 50 and 100 tags across different customer environments and use cases. The company has begun generating commercial traction, with revenues currently in the ₹3 million to ₹5 million range. Looking ahead, we aim to reach approximately US$1 million in revenue by 2027 and US$4 million to US$5 million by 2028.
Q. How does smartphone UWB affect your roadmap? Can smartphones replace tags?
A. Smartphones will expand the ecosystem, but they will not replace dedicated industrial tags. Industrial applications require rugged, ultra-small, low-power, always-on devices with long battery life and predictable performance. Smartphones, with their sensors, connectivity and computing capability, will complement the platform by supporting worker interaction, identification and selected tracking applications. The future will be a hybrid ecosystem in which smartphones, wearables and dedicated tags coexist, each performing the roles for which they are best suited.
Q. How are you balancing edge AI and cloud AI for industrial deployments?
A. Today, large enterprises generally prefer edge deployments because of stringent security requirements and the need for real-time, low-latency decision-making, with sensitive operational data often remaining on local infrastructure.
Over time, we expect a hybrid architecture to become the norm. The edge will handle real-time processing and mission-critical operations, while the cloud will support long-term analytics, business intelligence, AI training, retrieval-augmented generation (RAG) and large-scale data processing.
This approach combines the speed of edge computing with the scalability and analytical capabilities of the cloud.
Q. What will future AI-driven factory operations look like?
A. The future is not about humans competing with machines, but collaborating with them in a shared ecosystem. Today, workers still spend time on repetitive, low-value tasks such as constant scanning, which are better suited to automation.
Machines should handle routine processes, allowing people to focus on higher-value, decision-making activities. Our vision is a platform where humans, robots and autonomous systems work seamlessly together, with an AI layer orchestrating task allocation, optimising workflows and enabling continuous feedback between people and machines. The result is a more efficient, intelligent and collaborative industrial environment.
Q. What are the biggest challenges ahead?
A. One of the biggest challenges is that Physical AI is still an emerging technology, making customer education and validation essential. Organisations need to understand both its capabilities and its real-world value before deploying it at scale.
Another challenge is enabling intelligent collaboration between humans, robots and machines. As factories evolve towards mixed human-autonomous environments, our goal is to build an intelligence layer that connects them seamlessly, enabling coordinated interaction and a more efficient, collaborative industrial ecosystem.
Q. What is the long-term vision: product, platform or infrastructure layer?
A. We see ourselves as a platform company rather than a product company. Products solve individual problems, whereas platforms create ecosystems.
Our first step is to digitise the physical world by making assets, people and activities visible in real time. On this foundation, AI layers can generate insights, automate operations and optimise decision-making.
Our long-term vision is to build a complete Physical AI platform that delivers real-time data, analytics, human-machine orchestration and autonomous operational intelligence, becoming the foundational intelligence layer for AI-driven industrial operations.
