Most devices, such as cars, drones, and chips, function thanks to software that remains invisible to the general public—these are low-level drivers and firmware closely tied to the hardware. Creating such code is a labor-intensive and expensive process that still heavily relies on manual checks and documentation required for compliance. Furthermore, when migrating a product to a new chip, this can take a lead engineer months.
General artificial intelligence coding assistants, which are tuned for web applications and software, cannot provide significant help in this area because they lack an understanding of the processor's register map or its timing constraints. The startup H2LooP from Bengaluru is developing exactly this kind of artificial intelligence.
The company was founded in May 2025 by Sairanjan Mishra and Pulkit Agrawal. Both previously had experience building companies. Mishra, who serves as CEO, spent over fifteen years developing software for complex systems at Philips, Toshiba, Cisco, and Bosch, and is well aware of the slow procurement pace in these industries. Agrawal, the CTO, spent a decade at Google working on low-level software.
According to Mishra, in such systems, writing the code itself is rarely the main difficulty. Instead, teams spend months understanding the code, figuring out how it interacts with the hardware, and proving its correctness.
The H2LooP solution is based on a combination of two elements. First, it is a set of small language models specifically trained on embedded systems code, not on the general internet text content used by most models. Second, it is a patented knowledge graph that unifies hardware specifications, safety standards, design patterns, and the client's proprietary codebase into a single map that the models can analyze.
The main goal of this combination is to ensure context. Artificial intelligence writing a device driver without understanding the target chip will create code that looks correct externally but will fail upon integration. A verification layer then checks whether the generated code is correct and auditable before it goes into production. The entire system is capable of operating locally or in a completely isolated environment, ensuring that the client's code never leaves their control, which is critically important for the defense and semiconductor industries.
H2LooP sells its solutions to regulated, mission-critical industries: semiconductors, automotive, aerospace, telecommunications, and defense, as well as increasingly data centers and robotics. The company has already implemented its tools in semiconductor and automotive industry teams across India and Europe.
In one German semiconductor company dealing with legacy code conversion and AUTOSAR standard compliance, the company stated that its tools reduced design and audit time while generating test scenarios with an accuracy of 90% to 95%. Overall, all implementations report approximately a doubling of engineering speed. Additionally, H2LooP was selected to participate in the Infineon Startup Program and recognized by the Indian Electronics and Semiconductor Association.
In early 2026, H2LooP raised $2 million in a seed round co-led by Speciale Invest and 3one4 Capital. These funds are intended to strengthen the platform, scale enterprise deployments, and enter more complex areas such as data centers, drones, and robotics. The funding came at a favorable time: investments in Indian AI have sharply increased during 2025 and 2026, and companies manufacturing equipment for automotive, chip, and defense sectors face the same software challenges that H2LooP solves. The founders put forward a national argument: India has long created some of the world's most critical software, and now they aim to create an AI level that makes this software faster and more reliable.
Mowito, a company specializing in physical artificial intelligence (AI) and creating foundational models for industrial robots, announced on Tuesday that it has raised $3 million in a seed funding round.
The round was led by Version One Ventures. Other participants included All In Capital, Unisol, iSeed, as well as angel investors such as Soumith Chintala (from Thinking Machines Lab), Adarsh Kulkarni (from Foundry Robotics), Ashish Kulkarni (from Coformer.ai), and Vaibhav Domkundwar (from Better Capital).
The company stated that the funds will be used to accelerate expansion in the United States, strengthen engineering and market teams, and scale deployments with manufacturers in the automotive and electronics industries.
Founded in 2024, Mowito develops physical AI models that allow industrial robots to learn directly from task demonstrations. This eliminates the need for traditional programming while maintaining the necessary precision for manufacturing processes.
The company serves global manufacturers in the automotive and electronics sectors, with headquarters in Bengaluru and Detroit. It was noted that Mowito robots are already being used on production lines of a major Fortune 500 automotive company and one of the world's largest electronics contract manufacturers, supporting high-precision assembly operations.
Co-founder and CEO of Mowito, Puru Rastog, emphasized that manufacturing has reached a point where the bottleneck is no longer hardware but software. He noted that robots should not require reprogramming when production changes, but should learn like humans—through observation and repetition. He added that this funding will help accelerate this concept, expand globally, and implement Physical AI in more manufacturing environments.
All In Capital Partner Kushal Bhagia expressed the view that manufacturing is entering a new phase where AI will radically change industrial automation. He highly praised Mowito, stating that the company is creating fundamental technology that removes one of the main barriers to industrial automation—the complexity of robot programming—and that the technical depth of the team, early customer validation, and vision for Physical AI place it in an exceptional position to define this category.
