Real-Time Machine Learning | 19 566

Opportunity Information: Apply for 19 566

The Real-Time Machine Learning (RTML) grant opportunity is aimed at pushing machine learning beyond today s mostly offline, batch training workflows and into systems that can learn continuously from streaming data while operating under tight time and energy limits. The core problem it targets is a long-standing challenge in computing: building machines that can interpret incoming information as it arrives, update what they know immediately, use that updated knowledge to handle unfamiliar situations, and do all of this with energy efficiency closer to the human brain than to power-hungry modern compute platforms. The opportunity frames this as essential for next-generation applications where waiting to retrain a model later is not acceptable, including autonomous vehicles, defense and military systems, healthcare informatics, and business analytics. In these settings, data arrives nonstop, conditions change quickly, and decisions often need to be made on the spot, which forces learning systems to be both fast and efficient.

A key emphasis of the solicitation is that real-time learning is not just a software problem; it requires new thinking in both machine-learning methods and the underlying hardware that runs them. While recent years have produced impressive large-scale machine learning results and specialized electronic hardware that supports heavy computation for tasks like speech recognition and computer vision, the solicitation argues that emerging domains look different: they involve large, continuously streaming datasets and operational constraints that make conventional approaches too slow, too power intensive, or too reliant on periodic retraining. As a result, the program calls for novel hardware techniques and new learning architectures designed specifically to support continuous, real-time processing and adaptation.

The technical focus is on co-design, meaning proposals are expected to treat algorithms and hardware as a coupled system rather than optimizing each in isolation. The program is looking for foundations for next-generation approaches where all stages of training can occur in real time, not only inference. That includes incremental training (updating models as new data arrives), hyperparameter estimation (tuning settings that often require expensive search procedures in traditional pipelines), and deployment (moving updated models into operation without long delays). In practical terms, the ideal outcomes would be architectures and learning methods that can keep learning after deployment, remain stable and reliable as conditions drift, and maintain high performance without requiring massive energy consumption or cloud-scale resources.

The opportunity is a joint effort between the National Science Foundation (NSF) and the Defense Advanced Research Projects Agency (DARPA). It is structured to encourage high-performance and energy-efficient solutions, and it highlights the potential for post-award collaboration between DARPA-supported and NSF-supported researchers. That detail matters because it signals an intent to bridge fundamental academic research (a traditional NSF strength) with mission-driven and hardware-forward development perspectives often associated with DARPA, especially for systems that must operate in demanding real-world environments.

From the published opportunity details, this is a discretionary grant solicitation from NSF under Funding Opportunity Number 19-566, created March 7, 2019, with an original closing date of June 6, 2019. The activity category is science and technology and other research and development, with CFDA numbers 47.041 and 47.070. The listing indicated an expectation of around 12 awards, and it did not specify an award ceiling in the summary field provided. Eligibility is listed broadly as "Others" with additional clarification referenced in the full text, implying applicants would need to consult the official solicitation for the exact eligible organization types and any constraints. Overall, the program is essentially a call for research that makes real-time, continuously learning machine intelligence practical by inventing new algorithm-and-hardware pairs that can train, adapt, and deploy on the fly under real operational constraints.

• The National Science Foundation in the science and technology and other research and development sector is offering a public funding opportunity titled "Real-Time Machine Learning" and is now available to receive applicants.
• Interested and eligible applicants and submit their applications by referencing the CFDA number(s): 47.041, 47.070.
• This funding opportunity was created on Mar 07, 2019.
• Applicants must submit their applications by Jun 06, 2019. (Agency may still review applications by suitable applicants for the remaining/unused allocated funding in 2026.)
• The number of recipients for this funding is limited to 12 candidate(s).
• Eligible applicants include: Others (see text field entitled Additional Information on Eligibility for clarification).

Apply for 19 566

[Watch] Creating a grant proposal using the step-by-step wizard inside the applicant portal: