Robotics has become central to modern manufacturing. From automobiles to electronics, semiconductors, logistics and warehouses, factories worldwide rely on robotic arms and machines.
The International Federation of Robotics (IFR) World Robotics 2024 report shows that the global stock of industrial robots reached 4.28 million in 2023, a 10% rise from the previous year. Asia leads, with China accounting for 51% of global installations at 2,76,288 units. India is catching up fast, installing a record 8,510 robots in 2023.
Key trends for 2025 include the rise of AI and humanoids, a push for sustainability, new models like robot-as-a-service (RaaS) and robotics to offset global labour shortages.
CynLr, a Bengaluru-based robotics startup, is building machines that can adapt to new tasks without pre-programming or specialised hardware. Founder Gokul NA said the goal is to strip away complexity from factories.
He noted that making robots work is still tedious. “They are not intuitive with objects, even less than what a human baby would be. What comes naturally and instinctively to humans does not come so for the robots.”
“So, our job is to make the manufacturing as boring as possible,” he joked in an interview with AIM. Unlike conventional robots that depend on precision alignment and extensive training, CynLr’s machines can pick, orient and place objects they have never seen before. The company plans to deploy its robots in factories by early next year and raise new funding.
Making Robots Self-Learning
CynLr claims this flexibility could make manufacturing cheaper and easier to reconfigure. Gokul compared their robots to infants learning through trial and error. “It has no idea what it has picked,” he said during a walkthrough of the company’s facility.
The system tries to grasp, drops the object and tries again. These “primitives” let the robot build autonomy without huge datasets or rigid programming. CynLr has also developed its own camera modules. They use autofocus and eye-like convergence to sense depth.
“You can never see an object without converging your eyes,” Gokul explained. This design eliminates the complexity and errors of depth cameras, allowing robots to handle bolts, wires and flexible parts without requiring engineers to fix their orientation first.
CynLr’s ‘Dumb Baby’ Robot
At the centre of CynLr’s work is CyRo, a robot built to rethink factory automation. Gokul said if robots can pick an object in its natural orientation and then figure out how to place it, tasks become simpler and faster.
Most factory automation today depends on custom-built machines for specific parts. “You have 70% of the cost into any automation today…entirely about those customisations that you’ve done in the environment,” he said.
CynLr wants factories to be reprogrammable, closer to how computers work. “Then, it becomes like a laptop, and you can write everything. It becomes a software-defined factory,” he said. Reprogramming a robot for a new task could then take minutes, not months.
CyRo uses real-time sensing and correction, behaving, as Gokul put it, “like a dumb baby…just figuring out the best orientation to pick it up.” This lets it adapt to errors, variability and dynamic environments where conventional robots fail.
Vision is Not Sight
Vision is key to CyRo’s adaptability. More than half of the human brain is devoted to visual processing, and CynLr aimed to replicate that in machines. Unlike conventional systems that rely on colour, CyRo perceives motion, orientation and context.
“Our vision is way more than what we see as colour; it sees motion first,” Gokul said. This allows CyRo to act on objects that appear different under varying lighting or orientation.
By combining vision with adaptive learning, CyRo moves closer to what CynLr calls ‘object sentience’—the ability to turn a stimulus into an action without exhaustive programming.
Jobs Will Not Be Displaced
The rise of adaptable robots often sparks concern about job losses. Gokul dismissed the fear. “Displacement of jobs is what we are hyper-focused on because that’s too evident for us. But that didn’t happen in history, in the past.”
According to him, roles will evolve. Workers will move from repetitive tasks to teaching robots and managing adaptive systems. “Who else is best equipped to teach the robot than the guy who actually did that job?” he asked.
Unlike the steep learning curve of the software era, training robots is intuitive, relying on demonstration and correction. “How much of a skill barrier do I have to face to adapt from here to there? I think that is going to be much more minimal compared to what you faced between the computer and software,” he said.
93% Components Still Imported
CynLr faces challenges beyond technology. “We don’t have just the task of building a company. We have a task of building an industry along with it,” Gokul said.
Today, 93% of hardware value is imported. Key parts like image sensors and encoders come from Japan, South Korea, Israel and Europe. “Government’s understanding of Make in India is a lot more sound, but create in India has never been focused,” he said.
He also pointed to the mismatch between venture capital cycles and the realities of deep tech. Software-friendly funding timelines are too short for robotics. Gokul called for “new investment vehicles with long-term horizons” to support foundational technologies.
CynLr sees applications across automotive assembly, semiconductors, electronics, warehouses, industrial kitchens and even space labs. The company employs over 60 people and expects to reach 80 this year.
For Gokul, the mission goes beyond building robots. It is about creating an ecosystem for India to lead in automation. “What computers did for data, we are trying to do for objects,” he said.
The post This Bengaluru Firm is Building Robots That Won’t Wait for Your Instructions appeared first on Analytics India Magazine.
