How CynLr’s Object Intelligence Stack heralds the dawn of software-defined factories

The stack reflects a paradigm that shifts training from datasets to emulating the active sensing and innate curiosity of a human infant

In the vast assembly floors of global manufacturers — from automakers to semiconductor behemoths, despite the ubiquity of robotic arms in promotional videos, the reality of industrial automation is surprisingly limited. In a typical car assembly plant, as much as 90 percent of tasks remain the preserve of human hands.

One vital missing piece is the inability of robots to handle unfamiliar objects or adapt to new situations without laborious, custom programming. However, a new architectural approach to physical AI suggests that the factory of the future may look less like a rigid monument to mass production and more like a versatile, software-defined computer.

Among the next-generation deep tech ventures out of India that are at the forefront of this transition is CynLr — short for Cybernetics Laboratory — a robotics company that has spent years developing what its founders Gokul NA and Nikhil Ramaswamy describe as an Object Intelligence Stack.

CynLr is based in Bengaluru, with an advanced research centre in Switzerland and feet-on-ground customer-facing operations in the US. The company serves a clientele of global manufacturing giants. Its Object Intelligence Stack — developed over a period of five years — is but one example of CynLr’s broader effort to shift robotics from being data-obsessed or form-factor-obsessed to a paradigm that relies on emulating the active sensing and intuitive curiosity of a human infant.

This translates to robotic arms with the ability to grasp objects not because they have been trained on millions of images and videos, but because they recognise the fundamental physics of the items in front of them.

This shift in technical architecture heralds a shift in industrial networks and supply chains. For more than half a century, the economic logic of manufacturing has dictated the creation of centralised behemoths that are today sometimes referred to as ‘giga-factories.’ They entail capital intensity that can run into billions of dollars.

Therefore, they must produce tens of thousands of units, like cars, for example, month after month, year after year just to break even. In an era of rapid design cycles and fickle consumer tastes, such rigidity is becoming a liability. Therefore, the factory itself is becoming the new product, Gokul says — a modular, reconfigurable system that can be updated via software like we do our smartphones.

The result may be the rise of micro-factories: fabrication facilities as small as a car dealership or a local garage, even, located closer to the end buyer. Such systems would allow for genuine customisation — for instance, a car seat tailored precisely to a customer’s physical dimensions — without the prohibitive costs associated with manual bespoke work.

For global manufacturers, this agility won’t be just about costs. It represents a shot at setting up multiple, hyperlocal market expansion centres. The micro-factories could help them retain and expand market share by responding to trends very quickly.

Perhaps the most significant long-term value of these autonomous systems lies in recycling. Currently, “consumer mining” — the extraction of raw materials from discarded electronics — is often unviable because the manual labour required to disassemble complex gadgets is too expensive.

If a robot possesses the physical intelligence to assemble a mobile phone, it can, with the right software instructions, disassemble it just as efficiently. By automating the effort of separation, the materials within an old device regain their value, potentially creating a closed-loop system where the raw materials for a new product come directly from its predecessor.

The journey toward this “manipulation OS” is probably a decade-long effort. Challenges remain, from the stabilisation of deep-tech supply chains to the retention of specialised talent. Nevertheless, the transition from rigid CNC machines to cyber-physical systems is well underway.

As manufacturing evolves, the focus is shifting from the hardware itself to the “object intelligence” that governs it, promising a future where the production of physical goods is as flexible and decentralised as the digital world.