Download PDF Fall Bridge on the Materials Genome Initiative September 29, 2025 Volume 55 Issue 3 The Fall 2025 issue explores the Materials Genome Initiative’s progress and future outlook, showcasing advances in autonomous experimentation, sustainable polymers, next-generation batteries, and the broader role of AI in engineering. Accelerating the Materials Genome Initiative with Self-Driving Labs Tuesday, September 30, 2025 Author: Milad Abolhasani Self-driving labs promise to turn the Materials Genome Initiative’s bold vision into reality. The Materials Genome Initiative (MGI), launched in 2011, set a national goal to discover, manufacture, and deploy new materials at twice the speed and half the cost of the status quo (MGI n.d.). While substantial progress has been made through computational methods and curated data infrastructures, experimental bottlenecks persist. Self-Driving Laboratories (SDLs) offer a transformative pathway to overcome this limitation. SDLs integrate robotics, artificial intelligence, autonomous experimentation, and digital provenance in a closed-loop system capable of rapid hypothesis generation, execution, and refinement in a closed-loop fashion. This article explores how SDLs can activate each layer of the MGI, from foundational data generation to workforce development, and presents a 10-year roadmap for integrating SDLs into a national Autonomous Materials Innovation Infrastructure. Introduction: From Genome to Genius The Human Genome Project redefined biology through coordinated, large-scale data generation, leading to breakthroughs in medicine and genetics (Watson 1990). MGI extends this model to materials science, aiming to reduce discovery timelines through the integration of computation, data, and experiment (de Pablo et al. 2019). However, despite theoretical and simulation advances, a critical barrier remains: empirical validation. Physical experimentation still relies on manual procedures, limited throughput, and fragmented infrastructure, hampering the pace of materials innovation. SDLs can close this gap. By combining programmable hardware with artificial intelligence (AI)-driven decision engines, SDLs create autonomous platforms capable of performing thousands of experiments with little or no human intervention (Abolhasani and Kumacheva 2023) (see Figure 1A). SDLs redefine experimentation as a continuous, data-rich, adaptive process (Delgado-Licona et al. 2025). They are not replacements for human intuition but instead serve as powerful robotic collaborators that can test and iterate on ideas far beyond what is practical in conventional settings (Canty et al. 2025). The implications of SDLs are profound. For example, a national network of SDLs continuously generating validated datasets for new battery chemistries, semiconductor heterostructures, or polymer formulations could reduce time-to-solution by 100´ to 1,000´ times compared to the status quo. These platforms could rapidly identify promising candidates, validate theoretical predictions, and flag anomalous behaviors worthy of deeper study (Bennett et al. 2024; Boiko et al. 2023; Dai et al. 2024; Epps et al. 2020; Koscher et al. 2023; MacLeod et al. 2020; Slattery et al. 2024; Snapp et al. 2024; Steiner et al. 2019; Szymanski et al. 2023; Volk et al. 2023). More than just tools, SDLs become infrastructure—an autonomous experimental layer in the materials research ecosystem. SDLs represent the missing experimental pillar of the MGI vision. Their integration would not only accelerate discovery but also enhance reproducibility, access, and resilience. By converting experimentation into a programmable, scalable infrastructure, SDLs can become essential assets in achieving the MGI’s objectives. This article outlines how SDLs complement MGI’s strategic pillars, demonstrates how they can be deployed at scale, and explores how their widespread adoption can transform both how and who conducts materials science. The Materials Genome Initiative: Vision and Strategic Evolution Since its inception, the MGI has driven coordination across federal agencies, national laboratories, academia, and industry. Early successes were primarily computational. Initiatives such as The Materials Project (Jain et al. 2013), Open Quantum Materials Database (OQMD) (Kirklin 2015), and the Automatic FLOW for Materials Discovery database (Curtarolo et al. 2012) provided researchers with instant access to millions of calculated material properties. These resources allowed researchers to screen candidate materials virtually, reducing the time and cost of identifying promising materials with intriguing properties. Despite these achievements, the experimental layer of MGI remains underdeveloped. Data generation is often manual, idiosyncratic, and difficult to scale. Experimental metadata are inconsistently recorded, making reproducibility and cross-laboratory validation challenging. More fundamentally, the speed at which computational models improve outpaces our ability to generate corresponding experimental datasets. This misalignment limits the feedback loop between computation and experimentation that is central to MGI’s vision. By combining programmable hardware with artificial intelligence (AI)-driven decision engines, SDLs create autonomous platforms capable of performing thousands of experiments with little or no human intervention. The 2021 and 2024 MGI strategic documents explicitly recognize this gap (MGI n.d.). Both reports call for autonomous systems that can generate high-quality, reproducible data in a scalable and shareable format. This includes automated synthesis as well as characterization and integration of AI to design, interpret, and adapt experiments. The term “Autonomous Materials Innovation Infrastructure” emerged as a framework to conceptualize this vision (MGI 2024a). SDLs are uniquely positioned to operationalize this mandate. By strengthening all three foundational pillars of MGI—computation, data infrastructure, and experimentation—SDLs provide an integrated environment where theory, simulation, and empirical validation converge. Furthermore, SDLs directly address high-priority 2024 MGI Challenge Areas (MGI 2024b) such as decarbonization, semiconductor manufacturing, and critical mineral recovery by enabling rapid, reproducible testing and optimization. With their closed-loop learning and real-time adaptability, SDLs bring us closer to a truly predictive materials science ecosystem. The Architecture and Algorithms of Self-Driving Labs At a technical level, an SDL consists of five interlocking layers (see Figure 1): Actuation Layer: Robotic systems that perform physical tasks such as dispensing, heating, mixing, and characterizing materials; Sensing Layer: Sensors and analytical instruments that capture real-time data on process and product properties; Control Layer: The software that orchestrates experimental sequences, ensuring synchronization, safety, and precision; Autonomy Layer: AI agents that plan experiments (decision making), interpret results, and update experimental strategies (model refinement); and Data Layer: Infrastructure for storing, managing, and sharing data, including metadata, uncertainty estimates, and provenance. The autonomy layer distinguishes SDLs from traditional automation. Rather than executing a fixed set of experiments, an SDL interprets results and decides what to do next. This is crucial for navigating complex, nonlinear, or poorly understood materials spaces. For instance, in optimizing catalytic activity, an SDL may shift focus from composition to temperature as more is learned about the system, mimicking a human researcher’s strategy. Algorithms such as Bayesian optimization and reinforcement learning allow SDLs to efficiently navigate complex, multidimensional design spaces (Abolhasani and Kumacheva 2023). Large language models further enhance SDLs by translating user intent from scientific literature or natural language prompts into structured experimental constraints (Boiko et al. 2023; Ruan et al. 2024). Recent advances in AI have further enhanced the autonomy layer of SDLs. Multi-objective optimization frameworks can balance trade-offs between conflicting goals such as cost, toxicity, and performance. Uncertainty-aware models ensure that the SDL explores areas where predictions are weak, reducing bias. Large language models can parse scientific literature and translate user intent into experimental constraints. These developments expand the capabilities of SDLs, making them not only faster but smarter. An exemplary SDL—an autonomous multiproperty-driven molecular discovery (AMMD) platform—is illustrated in Figure 2. This SDL unites generative design, retrosynthetic planning, robotic synthesis, and online analytics in a closed-loop format to accelerate the design-make-test-analyze (DMTA) cycle (Koscher et al. 2023). It iteratively proposes dye-like molecules optimized for targeted physicochemical properties, synthesizes them, measures their properties in real time, and retrains its models with the new data. AMMD autonomously discovered and synthesized 294 previously unknown dye-like molecules across three DMTA cycles. The platform showcases how an SDL can explore vast chemical spaces and converge on high-performance molecules with autonomous robotic experimentation. In practice, SDLs have already demonstrated remarkable results (Table 1). In quantum dot synthesis, SDLs have mapped compositional and process landscapes an order of magnitude faster than manual methods (Bateni et al. 2024; Epps et al. 2020). In polymer discovery, they have uncovered new structure–property relationships that were previously inaccessible to human researchers (Snapp et al. 2024; Wang et al. 2025). These examples underscore the transformative potential of SDLs in bridging the gap between computation and real-world experimentation that is a central goal of the MGI. SDL Deployment Models: Centralized, Distributed, and Hybrid Scaling SDLs to fulfill the MGI vision requires thoughtful deployment strategies. Two dominant models are emerging, Centralized SDL Foundries and Distributed Modular Networks (Canty et al. 2025). Centralized SDL Foundries concentrate advanced capabilities in national labs or consortia. These facilities can host high-end robotics, hazardous materials infrastructure, and specialized characterization tools. They offer economies of scale and can serve as national testbeds for benchmarking, standardization, and training. Researchers can submit digital workflows to be executed remotely, facilitating access to cutting-edge experimentation. In contrast, Distributed SDL Networks enable widespread access by deploying low-cost, modular platforms in individual laboratories. Though more modest in scope, these distributed SDL platforms offer flexibility, local ownership, and rapid iteration. When orchestrated via cloud platforms and harmonized metadata standards, they function as a “virtual foundry,” pooling experimental results and accelerating collective progress. A list of available SDL infrastructure across North America is presented in a 2024 MGI report (Subcommittee on the MGI 2024). A universal SDL model will offer the best of both worlds. Preliminary research can be conducted locally using distributed SDLs, while more complex tasks are escalated to centralized SDL facilities. This layered approach mirrors cloud computing, where local devices handle basic computation and data-intensive tasks are offloaded to data centers. In the SDL context, such a model can maximize both efficiency and accessibility. Deployment models must also consider interoperability, cybersecurity, and sustainability. Interoperable SDLs require open application programing interfaces (APIs), shared data ontologies, and robust orchestrators. Cybersecurity is critical given the physical risks associated with autonomous experimentation. Sustainability considerations, such as reagent use, waste generation, and energy consumption, must be integrated into the SDL design AND operation. These considerations shape how SDLs are built, governed, and integrated into national infrastructure to align with MGI’s long-term goals. Data, Autonomy, and the MGI Knowledge Loop At the core of SDL utility is their ability to generate, curate, and interpret data at unprecedented scales and speed. In traditional human-centered experimentation settings, data quality often varies significantly depending on instrumentation, operator expertise, and contextual documentation. SDLs resolve these inconsistencies by encoding every step of the experimental process into machine-readable records. This digitalization includes reagent identities and volumes, as well as equipment settings, environmental conditions, and calibration metadata. Autonomous agents operating within SDLs are capable of managing and optimizing the computational and experimental data lifecycle. When orchestrated correctly, the SDL can identify correlations and causal links that may be opaque to human interpretation, particularly in high-dimensional design spaces. For example, in exploring the composition–process–property space of multi-cation oxides, an SDL can link trace impurity levels or subtle thermal gradients to variations in material functionality, thereby enabling fine-tuned control on material synthesis, composition, and properties that would otherwise require years of labor-intensive study. Furthermore, SDLs are natural engines for active learning. Their ability to use real-time feedback to refine predictive models means they can operate efficiently even in data-sparse regimes. Instead of brute-force sampling, SDLs prioritize experiments that maximize information gain. This strategy is particularly valuable in systems with combinatorially large variable spaces, such as doped semiconductors or hybrid organic–inorganic materials, where exhaustive sampling is computationally and experimentally intractable. A universal SDL model will offer the best of both worlds. SDLs also improve the feedback loop between simulation and experiment. When integrated with multiscale modeling strategies, SDLs can validate and refine simulations continuously, tightening the predictive cycle (Gongora et al. 2021). By leveraging proxy measurements and uncertainty quantification, SDLs can infer properties not directly observable in the lab (Osterrieder et al. 2023). This integration is core to MGI’s mission of accelerating discovery through coupled computation and experiment. In addition to enabling deeper understanding of materials relationships, SDLs also facilitate rigorous data standardization through embedded provenance protocols. Each SDL-generated dataset can be traced back through its full experimental history, including sensor calibrations, environmental logs, reagent batch numbers, and algorithm version. This capability supports reproducibility and meta-analyses, as well as integration with MGI-curated repositories such as The Materials Project, allowing researchers to explore how small differences in conditions or instrumentation affect outcomes. Such high-resolution traceability is foundational for integrating experimental data into machine-readable knowledge graphs that power future hypothesis generation. One of the most powerful use cases of SDL-generated data is transfer learning, leveraging models trained in one context to make predictions in another. A library of bandgap measurements in one class of semiconductors, for example, may help inform the optimization of a related class with limited prior data. Transfer learning is only effective when datasets are standardized, well-documented, and interoperable. The promise here is not only accelerating new discoveries but also improving generalization and reusability of scientific knowledge. These capabilities close the loop on MGI’s core vision: a seamless knowledge cycle from simulation to synthesis to deployment. Workforce and Ecosystem Evolution The SDL-driven transformation of experimental science demands a parallel evolution in workforce development and academic structures. SDLs blur disciplinary boundaries, requiring fluency in robotics, automation, programming, analytical chemistry, and data science. To fully realize the SDL-powered MGI vision, educational institutions must modernize training pipelines to prepare a new generation of interdisciplinary researchers. Within the SDL ecosystem, three complementary roles emerge: SDL Developers, responsible for designing and integrating the hardware–software stack; SDL Technicians, who maintain systems, calibrate instruments, and ensure operational robustness; and SDL Users, domain experts who frame scientific hypotheses, evaluate outputs, and interface with autonomy agents. Developing this tiered skill architecture necessitates new curricula, including interdisciplinary degree programs and hands-on modules with open-source SDL kits (Canty et al. 2025). Moreover, this evolution presents an opportunity to maximize accessibility and participation in advanced materials research. By lowering the skill barrier required to run complex experiments, SDLs can enable broader participation from all institutions and regions. With appropriate cloud infrastructure and remote interfaces, researchers can design and oversee experiments on a national SDL facility, gaining access to tools and data previously limited to few laboratories. Developing a robust talent pipeline will also require partnerships with industry and government. Internships and co-op positions at SDL facilities can provide hands-on experience while aligning training with real-world needs. Certification pathways, akin to those in welding or CNC machining, may be introduced for SDL operation, maintenance, and programming. Such credentials would validate workforce readiness and provide upward mobility for technical staff across academia and industry. Another key component of the SDL workforce ecosystem is the integration of social sciences and ethics. As automation reshapes research dynamics, scholars in science and technology studies, education, and public policy will be essential to monitor impacts, anticipate unintended consequences, and guide appropriate implementation. Embedding these perspectives early into SDL ecosystems will maintain a culture of responsible innovation. Partnerships and Economic Impact The economic case for SDL deployment is compelling. In industry, where R&D timelines, time-to-solution, and costs are tightly coupled to competitiveness, the advantages of SDLs are immediate. Autonomous workflows shorten design cycles, improve reproducibility, and reduce experimental waste. For instance, in pharmaceutical development, SDLs have already demonstrated their ability to identify optimal reaction conditions with an order-of-magnitude fewer experiments than the conventional design of experiments approaches. Academia–industry partnerships are pivotal for translating SDL capabilities into commercial impact (Bennett et al. 2024). National labs can serve as pre-competitive testbeds, industry can supply relevant use cases and application constraints; academia can push the frontier of AI, robotics, and experimental automation. Public–private consortia can share infrastructure costs while accelerating the validation and adoption of SDL workflows. Moreover, SDLs represent a novel paradigm for innovation-driven entrepreneurship by decoupling access to advanced experimental capabilities from direct ownership of capital-intensive infrastructure. This structural shift enables startups and small enterprises to engage in high-impact research and development (R&D) with significantly reduced upfront investment, thereby accelerating design–build–test–learn cycles and facilitating rapid scaling. An analogous model already exists in the pharmaceutical sector through contract research organizations (CROs), which provide specialized R&D capabilities and infrastructure on a service basis to clients lacking in-house resources. While CROs largely emulate the outsourced automation and high-throughput experimentation aspects of the SDL model, SDLs, with integrated AI-assisted decision making, advance this concept further by enabling closed-loop, hypothesis-driven experimentation that adapts in real time to emerging results. The established CRO ecosystem offers a useful reference point for understanding how a mature SDL ecosystem could function, while also highlighting the potential for SDLs to extend such shared-access models beyond pharmaceuticals to a broader spectrum of scientific and engineering domains. Growing venture capital investment in laboratory automation, particularly in biotechnology and cleantech, reflects this momentum. In this framework, SDLs shift the locus of value creation from physical experimental throughput to idea throughput, wherein competitive advantage derives from the development of algorithms, data-driven models, and optimized decision-making strategies rather than sheer experimental capacity. At the regional level, SDL infrastructure investment can anchor technology clusters, attracting talent, catalyzing spinoffs, and driving economic development. SDL-enabled hubs focused on semiconductors, green chemistry, or battery materials could mirror the role of semiconductor fabs and genomics centers in earlier eras. Policymakers should recognize SDLs as a strategic asset for national and regional innovation competitiveness. SDLs also create new economic value chains. Companies can build businesses around SDL software stacks and API standards. Vendors of analytical hardware are incentivized to produce SDL-compatible tools with modular designs. Even reagent suppliers may adapt to offer SDL-optimized consumables with QR-coded metadata for seamless integration. As these markets mature, the SDL economy will become a key segment of the broader materials innovation infrastructure. From a public-sector perspective, SDL investments can also enhance US competitiveness in critical technology domains such as clean energy, semiconductors, and pharmaceuticals. Federal funding for SDL research could be tied to national strategic initiatives, such as CHIPS and Science Act priorities or decarbonization roadmaps, aligning infrastructure development with urgent societal needs. A national SDL backbone can thus become a vital lever for mission-driven research and economic policy. Trust, Transparency, and Ethical Safeguards As with any powerful technology, the rise of SDLs brings ethical and philosophical challenges. Building trust requires transparency, reproducibility, and community engagement. The opacity of AI decision making can obscure how experimental strategies are chosen or interpreted. Black-box algorithms can lead to results that are hard to validate or explain. To address these critical matters, the SDL community must prioritize documentation and standardization. Every experiment must be traceable, with rich metadata that allows others to reproduce and interpret results. Benchmarking protocols should be established, allowing performance comparisons across SDL platforms. SDLs must report all data, uncertainty bounds, and operational constraints. Equally important, SDLs must be applied in a transparent and deliberate manner, ensuring that automated decision making aligns with clearly defined scientific objectives and ethical considerations. Security is also paramount for SDLs. Because SDLs interface with physical matter, software vulnerabilities can have real-world consequences. Protocols for cybersecurity, access control and fail-safe mechanisms must be built into every SDL system. As SDLs become more networked, the risk of tampering increase. Regulatory frameworks may need to evolve to account for these hybrid digital–physical systems. Facile access to SDL infrastructure should also be a priority. Strategies such as open-source platforms, shared facilities, and subsidized cloud access can facilitate access to SDLs. By designing for openness and equity from the outset, the SDL community can ensure that the benefits of autonomy are broadly shared. By designing for openness and equity from the outset, the SDL community can ensure that the benefits of autonomy are broadly shared. An additional dimension of trust-building involves the human-machine interface. For SDLs to be accepted by the broader scientific community, their decision processes must be interpretable. This need calls for explainable AI techniques that can articulate why an agent selected a particular experiment or rejected a hypothesis. Visual dashboards, natural language logs, and confidence metrics can help bridge the gap between algorithmic reasoning and human intuition. The establishment of SDL ethics boards may also prove beneficial. These interdisciplinary committees could review autonomous workflows for safety, access, and scientific validity, especially for high-risk applications. Similar to Institutional Review Boards (IRBs) in biomedical research, SDL ethics boards would promote transparency and accountability without impeding innovation. Public engagement will further enhance legitimacy of SDLs. Communicating how SDLs work, what data they collect, and how they make decisions can demystify the technology. Open lab days, digital twins for citizen science, or interactive dashboards could bring SDLs into classrooms and communities, showcasing their role in accelerating sustainable innovation. Strategic Recommendations To fully integrate SDLs into the MGI and national research infrastructure, a coordinated roadmap is essential. In the next 1–3 years, the focus should be on establishing standards and addressing the current SDLs’ engineering bottlenecks outlined in Table 2. Table 2 highlights that SDL maturity is uneven across different disciplines: while fluidic and thin-film chemistries are already automated, domains demanding extreme environments or nanometer precision still face steep technological barriers. For high-temperature alloys and ceramics, the bottleneck is robotic hardware survivability at high temperatures and the lack of in-situ phase probes. Heterogeneous catalysis and microelectronics similarly require purpose-built infrastructure (modular, safety-aware reactor skids and fab-class, multi-modal metrology) plus interoperable data standards to translate SDL principles into fully functional platforms. Other strategic recommendations include: Establishing a national SDL user facility, analogous to supercomputing centers, to provide access and build community expertise. Deploying federated SDL networks within 4–7 years to enable collaborative experimentation across institutions (Figure 3) and formalizing interoperability frameworks to allow seamless exchange of workflows and data. Scaling workforce development programs, including certifications and interdisciplinary degrees, to meet growing SDL demand. Aligning SDL deployment with key MGI priority areas (e.g., semiconductor resilience, clean energy materials, circular economy). Fully integrating SDLs into the national R&D ecosystem by year 10. Ensuring every major university hosts at least one SDL node for AI-assisted scientific research and workforce training. Coordinating shared repositories of workflows and data through national laboratories, and incorporate SDL pipelines into standard product development lifecycles across industry. These initiatives would help autonomous experimentation evolve from a niche capability into a foundational element of science and engineering. Another near-term priority is the development of common testing suites for SDL benchmarking. These benchmarks could include standard synthesis targets, characterization routines, and comparison metrics for evaluating performance across different SDL platforms. Such standardized testbeds would enable comparative studies, support best-practices documentation, and create a virtuous cycle of performance improvement. As SDL networks grow, federated learning architectures will become essential. Rather than centralizing all data, SDL nodes could train local models and share only aggregated updates, protecting intellectual property while improving global model accuracy. This architecture mirrors developments in healthcare AI and could be adapted to materials science with appropriate data governance protocols. By the end of the decade, SDLs could also be integrated into autonomous manufacturing ecosystems. For example, in the next decade, an SDL can design a polymer, send its formulation to an additive manufacturing module, evaluates the specific part’s mechanical properties, and feed back the data to optimize performance, all with minimal human intervention. This convergence of digital design, experimentation, and production would mark a new era of agile, intelligent manufacturing. Conclusion SDLs—a convergence of AI, robotics, and experimental science—are at an inflection point for how we generate scientific knowledge. For the MGI, SDLs offer a tangible path to bridge computation and reality, turning predictive models into validated materials with unprecedented speed. But their true impact lies beyond throughput or efficiency. SDLs can reshape who participates in science, how collaboration happens, and what becomes possible when experimentation itself becomes a programmable, adaptive system. To realize this potential, we must act with vision and coordination. Investment in SDL infrastructure is an investment in the next era of scientific discovery. The MGI began as a bold promise; SDLs are how we fulfill it. The SDL vision aligns with the scientific ambitions of the MGI and broader societal aspirations: reducing environmental impact, localizing manufacturing, accelerating medicine development, and ensuring rapid and facile access to innovation. It is a vision in which science becomes more dynamic and capable of meeting 21st century challenges. Investment in SDLs is an investment in the next phase of the MGI. SDLs activate the experimental pillar of the initiative, enable cross-domain data generation, and promote reproducibility and agility. SDLs support both national competitiveness and scientific inclusion. To seize this opportunity, leadership and coordination are essential. Federal agencies must fund and coordinate SDL infrastructure. Universities must rethink curricula and facilities. Industry must invest and co-develop SDLs with academia. And the scientific community must embrace new modes of discovery. With foresight and commitment, the SDL-powered future of the MGI is within reach, and its realization will mark a generational advance in the practice and promise of science. References Abolhasani M, Kumacheva E. 2023. The rise of self-driving labs in chemical and materials sciences. Nature Synthesis 2:483–92. Bateni F, Sadeghi S, Orouji N, Bennett JA, Punati VS, Stark C, Wang J, Rosko MC, Chen O, Castellano FN, and 2 others. 2023. Smart Dope: A self-driving fluidic lab for accelerated development of doped perovskite quantum dots. Advanced Energy Materials 14:2302303. Bennett JA, Orouji N, Khan M, Sadeghi S, Rodgers J, Abolhasani M. 2024. Autonomous reaction Pareto-front mapping with a self-driving catalysis laboratory. Nature Chemical Engineering 1:240–50. Boiko DA, MacKnight R, Kline B, Gomes G. 2023. Autonomous chemical research with large language models. Nature 624:570–78. Canty RB, Bennett JA, Brown KA, Buonassisi T, Kalinin SV, Kitchin JR, Maruyama B, Moore RG, Schrier J, Seifrid M, and 3 others. 2025. Science acceleration and accessibility with self-driving labs. Nature Communications 16:3856. Curtarolo S, Setyawan W, Wang S, Xue J, Yang K, Taylor RH, Nelson LJ, Hart GLW, Sanvito S, Buongiorno-Nardelli M, and 2 others. 2012. AFLOWLIB.ORG: A distributed materials properties repository from high-throughput ab initio calculations. Computational Materials Science 58:227–35. Dai T, Vijayakrishnan S, Szczypiński FT, and 8 others. 2024. Autonomous mobile robots for exploratory synthetic chemistry. Nature 635:890–97. de Pablo JJ, Jackson NE, Webb MA, Chen L-Q, Moore JE, Morgan D, Jacobs R, Pollock T, Schlom DG, Toberer ES, and 14 others. 2019. New frontiers for the Materials Genome Initiative. npj Computational Materials 5:41. Delgado-Licona F, Addington D, Alsaiari A, Abolhasani M. 2025. Engineering principles for self-driving laboratories. Nature Chemical Engineering 2:277–80. Epps RW, Bowen MS, Volk AA, Abdel-Latif K, Han S, Reyes KG, Amassian A, Abolhasani M. 2020. Artificial Chemist: An autonomous quantum dot synthesis bot. Advanced Materials 32(30):e2001626. Online at doi.org/10.1002/adma.202001626. Gongora AE, Snapp KL, Whiting E, Riley P, Reyes KG, Morgan EF, Brown KA. 2021. Using simulation to accelerate autonomous experimentation: A case study using mechanics. iScience 24:102262. Jain A, Ong SP, Hautier G, Chen W, Richards WD, Dacek S, Cholia S, Gunter D, Skinner D, Ceder G, and 1 other. 2013. Commentary: The Materials Project: A materials genome approach to accelerating materials innovation. APL Materials 1:011002. Kirklin S, Saal J, Meredig B, Thompson A, Doak JW, Aykol M, Rühl S, Wolverton C. 2015. The Open Quantum Materials Database (OQMD): Assessing the accuracy of DFT formation energies. npj Computational Materials 1:15010. Koscher B, Canty RB, McDonald MA, Greenman KP, McGill CJ, Bilodeau CL, Jin W, Wu H, Vermeire FH, Jin B, and 8 others. 2023. Autonomous, multi-property-driven molecular discovery: From predictions to measurements and back. Science 282(6677). Online at www.science.org/doi/10.1126/science.adi1407. MacLeod BP, Parlane FGL, Morrissey TD, Häse F, Roch LM, Dettelbach KE, Moreira R, Yunker LPE, Rooney MB, Deeth JR, and 10 others. 2020. Self-driving laboratory for accelerated discovery of thin-film materials. Science Advances 6(20):eaaz8867. MacLeod BP, Parlane FGL, Rupnow CC, Dettelbach KE, Elliott MS, Morrissey TD, Haley TH, Proskurin O, Rooney MB, Taherimakhsousi N,and 7 others. 2022. A self-driving laboratory advances the Pareto front for material properties. Nature Communications 13:995. MGI [Materials Genome Initiative]. n.d. Materials Genome Initiative. Online at www.mgi.gov. MGI. 2024a. Accelerated Materials Experimentation Enabled by the Autonomous Innovation Initiative (AMII): A Workshop Report. Online at www.mgi.gov/autonomous-experimentation-materials-rd. MGI. 2024b. The 2024 MGI challenges. Online at www.mgi.gov/2024-materials-genome-initiative-mgi-challenges. Osterrieder T, Schmitt F, Lüer L, Wagner J, Heumüller T, Hauch J, Brabec CJ. 2023. Autonomous optimization of an organic solar cell in a 4-dimensional parameter space. Energy & Environmental Science 16:3984–93. Ruan Y, Lu C, Xu N, He Y, Chen Y, Zhang J, Xuan J, Pan J, Fang Q, Gao H, and 4 others. 2024. An automatic end-to-end chemical synthesis development platform powered by large language models. Nature Communications 15:10160. Slattery A, Wen Z, Tenblad P, Sanjosé-Orduna J, Pintossi D, den Hartog T, Noël T. 2024. Automated self-optimization, intensification, and scale-up of photocatalysis in flow. Science 383(6681):eadj1817. Snapp KL, Verdier B, Gongora AE, Silverman S, Adesiji AD, Morgan EF, Lawton TJ, Whiting E, Brown KA. 2024. Superlative mechanical energy absorbing efficiency discovered through self-driving lab-human partnership. Nature Communications 15:4290. Steiner S, Wolf J, Glatzel S, Andreou A, Granda JM, Keenan G, Hinkley T, Aragon-Camarasa G, Kitson PJ, Angelone D, and 1 other. 2019. Organic synthesis in a modular robotic system driven by a chemical programming language. Science 363(6423):eaav2211. Szymanski NJ, Rendy B, Fei Y, Kumar RE, He T, Milsted D, McDermott MJ, Gallant M, Cubuk ED, Merchant A, and 6 others. 2023. An autonomous laboratory for the accelerated synthesis of novel materials. Nature 624:86–91. Volk AA, Epps RW, Yonemoto DT, Masters BS, Castellano FN, Reyes KG, Abolhasani M. 2023. AlphaFlow: autonomous discovery and optimization of multi-step chemistry using a self-driven fluidic lab guided by reinforcement learning. Nature Communications 14:1403. Online at doi.org/10.1038/s41467-023-37139-y. Wang C, Kim YJ, Vriza A, Batra R, Baskaran A, Shan N, Li N, Darancet P, Ward L, Liu Y, and 7 others. 2025. Autonomous platform for solution processing of electronic polymers. Nature Communications 16:1498. Watson JD. 1990. The Human Genome Project: Past, present, and future. Science 248(4951):44–49. Yik JT, Hvarfner C, Sjölund J, Berg EJ, Zhang L. 2025. Accelerating aqueous electrolyte design with automated full-cell battery experimentation and Bayesian optimization. Cell Reports Physical Science 6:102548. About the Author:Milad Abolhasani is Alcoa Professor, Department of Chemical & Biomolecular Engineering, North Carolina State University