IIT Madras Alumnus Raises $9 Million for AI Startup

IIT Madras alumnus Akash Ramdas is turning years of academic research into a business focused on one of the semiconductor industry’s growing challenges: managing the heat generated by powerful computer chips. His startup, Discovered Materials, has raised $9 million in seed funding to use Artificial Intelligence to accelerate the discovery of new materials, according to TechCrunch.

The company is developing AI agents that help scientists identify promising materials for semiconductor applications. By combining AI-generated candidates with physics-based modelling, computational simulations and laboratory experiments, Discovered Materials aims to reduce the time required to discover materials with useful properties.

The startup’s work comes as advances in artificial intelligence increase demand for powerful computing hardware. As AI chips handle increasingly demanding workloads, managing heat has become an important consideration in semiconductor design. Materials that can help address thermal challenges could contribute to the development of more efficient and reliable electronic devices.

Ramdas brings a background in metallurgical and materials engineering, computational materials discovery and nanoscale electronic devices to the venture. After graduating from IIT Madras, he pursued advanced studies and doctoral research at Stanford University before moving from academic research to entrepreneurship.

How Discovered Materials Uses AI to Find New Semiconductor Materials

Discovered Materials is building AI agents designed to generate and evaluate potential materials for applications such as heat management in semiconductor chips. Rather than relying entirely on researchers to propose and examine candidates individually, the platform aims to automate parts of the early discovery process.

According to the supplied report, the startup uses AI models from Anthropic to generate possible materials. These candidates are then assessed using physics-based models and computational simulations to determine which ones appear promising enough to warrant further investigation.

This approach combines the speed of AI-assisted exploration with established scientific methods. AI can help researchers search a large space of possible material compositions, while simulations can help assess whether a candidate’s predicted properties justify further work.

Materials science presents a particularly demanding discovery problem because relatively small changes in composition or structure can affect a material’s behaviour. Scientists must consider properties such as thermal performance, stability and compatibility with the intended application. Finding a candidate that performs well across the necessary criteria can require extensive experimentation.

AI-assisted discovery aims to narrow the search by identifying candidates that appear promising before researchers invest time and resources in producing and testing them. The technology does not eliminate the need for scientific validation, but it could help laboratories focus their efforts on a more targeted selection of materials.

AI Agents Can Explore Thousands of Materials a Day

One of the key differences highlighted by Discovered Materials is the scale at which its AI agents can explore possible candidates.

The supplied report says the system can examine thousands of potential materials in a day. During his PhD research, Ramdas explored approximately 20 candidates daily, providing a comparison between the pace of his earlier research and the scale the startup aims to achieve through automation.

The comparison illustrates the potential advantage of using AI to assist with repetitive computational tasks. A system capable of generating and screening large numbers of candidates could allow researchers to investigate a broader range of possibilities without manually performing every early-stage step.

However, the number of candidates explored is not the same as the number of materials successfully discovered or ready for commercial use. Generating a potential composition is only one part of the process. Researchers must still determine whether its predicted properties hold up under more rigorous evaluation and whether the material can be produced reliably.

For that reason, the practical value of AI materials discovery depends on more than speed. The quality of the predictions, the accuracy of the screening methods and the ability to validate promising candidates will all influence whether the technology produces useful results.

Discovered Materials is attempting to address this challenge by combining automated discovery with computational evaluation and experimental testing. Its approach seeks to make the early stages of research more efficient while retaining the scientific checks needed to establish whether a material works.

Why Semiconductor Chips Need Better Heat-Management Materials

Heat management is an increasingly important engineering challenge as semiconductor devices support more demanding computing tasks. AI systems rely on processors and accelerators that perform large volumes of calculations, and the resulting heat must be managed to keep hardware operating within its design limits.

Excessive heat can create difficulties for electronic systems by affecting performance, reliability and the conditions under which components can operate. Cooling systems and chip packaging therefore play important roles in the design of computing infrastructure.

Materials are central to these efforts because they influence how heat moves through and away from electronic components. Depending on the application, engineers may need materials that provide suitable thermal properties while also meeting requirements for electrical behaviour, structural stability and compatibility with semiconductor manufacturing processes.

Developing materials that satisfy multiple requirements can be difficult. A candidate with an attractive property in a simulation may not perform as expected when manufactured, integrated into a device or exposed to real operating conditions.

This creates an opportunity for AI-assisted materials discovery. By helping researchers screen a wider range of possibilities, AI tools could make it easier to identify candidates that merit detailed investigation for specialised semiconductor applications.

Nevertheless, the potential contribution of Discovered Materials should be distinguished from a proven improvement in commercial chip performance. The supplied report describes the startup’s technology and its goals, but it does not establish that its AI-discovered materials are already being widely used in commercially available semiconductors.

Discovered Materials Raises $9 Million in Seed Funding

Discovered Materials has raised $9 million in seed funding to advance its AI-driven materials discovery work, according to TechCrunch. The funding gives the startup an opportunity to develop its technology and pursue its goal of accelerating research into materials for semiconductor applications.

Seed funding is an early stage of startup financing, typically used to help a company develop its product, build its technical capabilities and establish a foundation for future growth. For a company working at the intersection of AI and materials science, that work can involve developing computational systems, improving prediction methods and connecting digital discovery with laboratory validation.

Materials discovery differs from many software applications because a promising computational result must ultimately be tested in the physical world. Even when AI identifies a candidate with attractive predicted properties, researchers still need to determine whether the material can be synthesised, whether it performs as expected and whether it can be manufactured consistently.

These requirements can make the route from an AI prediction to a commercially useful material lengthy and technically demanding. The funding could support Discovered Materials as it works to develop a process that links computational discovery with practical testing, although the supplied report does not detail how the company plans to allocate the full amount.

The $9 million raise also places the startup within a broader effort to apply AI to scientific research. Companies and research teams are exploring ways to use machine learning to identify promising candidates in areas where conventional discovery can involve large numbers of calculations and experiments.

For Discovered Materials, the central challenge will be demonstrating that its approach can produce validated materials that address real engineering needs, rather than simply increasing the number of candidates generated by AI.

Startup Releases AI-Discovered Materials and a Benchmark

Discovered Materials has also released hundreds of AI-discovered materials and introduced a benchmark designed to evaluate how well AI systems identify promising semiconductor materials, according to the supplied report.

Releasing candidate materials can give researchers additional information to examine and assess. A benchmark, meanwhile, can provide a structured way to compare the performance of AI systems on a defined discovery task.

Evaluation is important because an AI model may produce plausible-looking candidates without consistently identifying materials that meet a particular scientific objective. A benchmark can help researchers examine the quality of a system’s predictions and compare different approaches under specified conditions.

However, the usefulness of any benchmark depends on how it is designed, what it measures and how well its results reflect real-world scientific challenges. Strong performance on a computational test does not automatically establish that a system can produce a commercially viable material.

Similarly, the release of hundreds of AI-discovered candidates should not be interpreted as proof that all of them have been experimentally validated or are ready for industrial use. The supplied report does not establish that every released material has completed laboratory testing or demonstrated scalable manufacturing potential.

The distinction between generating candidates and validating them is especially important in materials science. A material may appear promising in a computational model but prove difficult to synthesise, unstable under operating conditions or unsuitable for integration into a finished device.

Discovered Materials’ progress will therefore depend on whether its computational methods can consistently guide researchers towards materials that survive these later stages of evaluation.

Laboratory Testing Remains a Major Challenge

Despite the potential of AI-assisted discovery, laboratory experiments remain a crucial step in determining whether a new material can deliver its predicted performance.

Computational models can estimate properties and help rank candidates, but they cannot replace every aspect of physical testing. Researchers must produce the material, measure its actual characteristics and determine whether it behaves as expected under relevant conditions.

Manufacturing at scale introduces additional questions. A material that can be produced in a research laboratory may not be economical or consistent enough for industrial use. Semiconductor applications can also impose strict requirements on purity, reliability and compatibility with existing production processes.

These barriers help explain why faster candidate generation does not necessarily translate into faster commercial adoption. Even if AI reduces the time required to identify a promising composition, subsequent experimentation, optimisation and manufacturing development may still take considerable effort.

According to the supplied report, AI-discovered materials have yet to achieve widespread commercial use, highlighting the gap between computational promise and industrial deployment.

For Discovered Materials, progress will depend on closing that gap. Its approach must show that AI-generated candidates can withstand experimental scrutiny and offer properties that matter for actual semiconductor applications.

Successful validation could strengthen the case for using AI as a practical research tool. Until then, the technology is best understood as a way to accelerate and focus materials exploration rather than as a replacement for experimental science.

From IIT Madras to Stanford University

Akash Ramdas’ academic journey began at the Indian Institute of Technology Madras, where he completed a Bachelor of Technology degree in Metallurgical and Materials Engineering in 2019.

He then moved to Stanford University in the United States, earning a master’s degree in Materials Science and Engineering between 2019 and 2021. He continued his research at Stanford, completing a PhD in Materials Engineering in June 2025.

His doctoral work focused on computational methods for discovering materials for nanoscale electronic devices. This research area connects materials science with the development of small-scale electronic systems, where the properties of a material can influence device performance and functionality.

Computational materials research uses models and simulations to investigate how materials might behave before researchers commit to making and testing them. The methods can help scientists explore potential compositions and properties, identify promising candidates and direct experimental work towards questions that matter most.

This background is relevant to the approach being developed at Discovered Materials. The startup’s goal of combining AI-generated candidates with physics-based evaluation builds on the broader principle of using computational tools to make the search for useful materials more efficient.

Ramdas’ transition from doctoral research to entrepreneurship reflects an effort to apply scientific methods beyond the academic setting. Instead of using computational discovery solely as a research technique, the startup is seeking to build a platform that can help other scientists and organisations investigate materials more quickly.

From Stanford Postdoctoral Researcher to Startup Founder

After completing his PhD in June 2025, Ramdas continued at Stanford as a postdoctoral researcher from September 2025 to April 2026, according to the supplied report.

He co-founded Discovered Materials in March 2026, while his postdoctoral research period was still underway. The transition placed his research expertise at the centre of a company focused on applying AI to materials discovery.

The combination of scientific research and commercial development presents a different set of demands. Academic work can focus on investigating specific questions and advancing scientific understanding, while a startup must also develop a usable product, demonstrate its value to potential customers and build a sustainable business.

In materials discovery, those goals are closely connected. A platform must do more than generate interesting predictions; it needs to help researchers make decisions, reduce unnecessary work or identify candidates that might otherwise be overlooked.

That makes experimental validation and practical utility important measures of the company’s progress. Its longer-term prospects will depend on whether the technology can deliver dependable results and fit into the research and development processes used by scientists and semiconductor companies.

What Discovered Materials’ AI Approach Could Mean for Scientific Research

Discovered Materials is part of a wider effort to use artificial intelligence to support scientific discovery. In fields where researchers must evaluate large numbers of possible candidates, AI can help automate parts of the search and prioritise options for closer examination.

Materials science is a natural area for this approach because the number of possible compositions and structures can be extensive. Searching that space entirely through conventional experimentation can require substantial time and resources.

AI systems may help scientists explore more possibilities, identify patterns in available data and select candidates for computational or laboratory evaluation. The greatest benefit is likely to come when these systems complement scientific expertise rather than operate without independent verification.

For semiconductor materials, the commercial relevance of a discovery depends on whether the material solves a clearly defined problem and can meet the requirements of a real device. Heat management is one potential application, but any proposed material must be assessed against the technical and manufacturing demands of its intended use.

The $9 million seed funding gives Discovered Materials resources to continue developing its approach, while its released candidates and evaluation benchmark provide additional elements of its research effort. The next challenge is to establish how effectively the platform can move from AI-generated possibilities to experimentally confirmed results.

Ramdas’ journey from IIT Madras to Stanford and then to a startup illustrates how a specialised academic background can become the foundation for a technology business. Whether Discovered Materials can turn its computational capabilities into materials that are adopted at scale will depend on scientific validation, manufacturability and its ability to meet the needs of the semiconductor industry.