Karthee Madasamy is an experienced deep-tech venture capitalist and the Managing Partner at MFV Partners, specializing in earlystage investments across the U.S., India, Israel and Southeast Asia. With a strong track record, he led successful investments in Waze (acquired by Google) and Validity Sensors (acquired by Synaptics). His expertise spans mobile technology, semiconductors, biometrics, cloud computing and AI-driven innovations. Madasamy brings operational experience from leading engineering and product marketing teams in multiple startups.
Meera Siva, CFA, is an early-stage startup investor with over 13 years of experience in finance and 16 years in engineering. She has managed the Shelter Venture Fund, supporting low-income housing startups, and led accelerator and grant programs. With expertise in due diligence, fundraising, and business model advisory, she has guided numerous startups. A former CFA Society India board member, Meera is also an experienced researcher and writer who has covered finance, startups, and social issues for a mainstream business newspaper. An electronic engineer by training, she holds multiple patents in microprocessors and networking.
Through this article, Madaswamy and Siva highlight the convergence of advancements in foundational technologies like sensors, software, and processing power, along with breakthroughs in AI/ML, as key enablers for significant progress and increased adoption of robotics, notably humanoid robots.
The internet was a huge technological advancement, but it was mobile technology that truly unlocked its power. Given how far mobile computing has come, it is worth asking where we are likely to see its magic shine next.
As someone who has worked in mobile technology at Qualcomm and has been an active participant in deep tech innovations (as a founder and investor), I am convinced that robotics is the next frontier, and it will see huge leaps in both advancement and adoption. Here’s why.
Building Blocks Exist
Robots have come a long way since their early industrial automation avatar. We were an early investor in Agility Robotics, and humanoids' abilities have improved phenomenally. Interestingly, this is not a surprise—the foundational technologies have been advancing steadily, and there have also been breakthroughs.
"The next step with AI and reinforcement learning methods will simulate demonstrations to enable tomorrow's robots to learn how to operate fully functional hands with joints and even fingertip capabilities"
For instance, one of the foundational technologies for robotics is sensor hardware. These have been following Moore’s law and becoming cheaper, faster and functionally rich in a small form factor. Additionally, software such as for vision systems has seen excellent research work and improvements that enable high precision and intelligence, besides being faster.
Likewise, at the systems level, integrating various functions has been made possible thanks to growth in processing power. Onboard processing has also been more reliable, requiring robots to integrate diverse sensor data (cameras, lidar, radar, IMUs) and use sophisticated algorithms to create a rich understanding of their environment.
AI/ML Advances
The biggest enabler, however, is the growth in Artificial Intelligence and Machine Learning in the recent few years. AI's ability to provide real-time learning, rapid simulation, and enhanced sensory inputs is enabling significant strides on many unsolved robotics problems, especially in the area of humanoid robots.
AI-driven simulations of the countless scenarios robots encounter are slashing development hours and costs. As mobility requires not just gathering and processing rich data from the environment but also understanding the context and adapting dynamically, intelligent learning models play a huge role in the value created by humanoids.
In recent years, humanoid robots have advanced from squeezing arms to lifting stuff, and today, you have a few fingers on some robots. The next step with AI and reinforcement learning methods will simulate demonstrations to enable tomorrow's robots to learn how to operate fully functional hands with joints and even fingertip capabilities.
Edge Computing
In the future, the key enabler for the shift from the current mobile model for robotics will be the ability to do AI on the edge. Rather than rely on cloud-based computing, the mobile device (robot) needs to perform significant computation for real-time decision-making. Machine learning models for tasks like perception, planning, and control must be run on the device. This is essential for autonomous navigation, object recognition, and adaptive behavior.
The computations must also be fast. Robots must make decisions and execute actions in real-time, so low-latency computation and robust control algorithms are needed.
The future is now
It is famously said that the future is already here but not uniformly distributed. This is very true for mobile robot platforms, as they are not something out there in the far future, but they already exist here and now.
Self-driving cars are one example of robots as mobile computing platforms. They require massive amounts of onboard processing to perceive the surroundings, make many driving decisions, handle unexpected surrounding conditions and control the vehicle. This must all be handled in real-time and quite reliably. As Waymo and others show, they have been becoming safer and more predictable, demonstrating the potential for the future of mobile robotics.
Industrial robots such as Digit from Agility Robotics can move autonomously, relying on sophisticated onboard processing to adapt to changing conditions and optimize their tasks. Humanoid robots will make computing truly mobile and be the platform where various verticals—not just industrial, retail, healthcare and many others—come together.
