United States

Agents for Navy Applications – A Practical Guide
Mark Tabladillo, Ph.D.
Cloud Solution Architect, Microsoft
Support Technical Advisor, Flank Speed
Abstract
This presentation explores the application of artificial intelligence (AI) agents in enhancing naval operations through the integration of Autogen and Azure AI Foundry. With the increasing complexity of naval missions and the rising threat of cyber-attacks, these tools provide a scalable and efficient solution for real-time decision-making and operational resilience.
The session highlights how hierarchical reinforcement learning (HRL) frameworks and modular agent architectures can be used to address the multi-level challenges of naval coordination. Autogen’s capabilities for creating and managing AI agents are combined with Azure AI Foundry’s robust infrastructure to demonstrate practical applications in ship coordination and cyber-attack mitigation.
A live demo showcases a simulated naval scenario where a fleet must navigate through a high-threat zone while responding to cyber-attacks. The demonstration illustrates how individual ship agents implement defensive measures, a command agent oversees fleet coordination, and large language models (LLMs) analyze intercepted communications for situational insights. These interactions highlight the agility and adaptability of AI systems in dynamic operational contexts.
The presentation concludes with a discussion on the scalability and future potential of Autogen and Azure AI Foundry in multi-domain naval operations, emphasizing their role in maintaining decision-making superiority and mission continuity in complex environments.
Citations:
Black, S. E. (2024). Mastering the digital art of war: Developing intelligent combat simulation agents for wargaming using hierarchical reinforcement learning (Doctoral dissertation). Naval Postgraduate School. Retrieved from https://hdl.handle.net/10945/73072
Black, S., &amp; Darken, C. (2023). Scaling intelligent agents in combat simulations for wargaming. Paper presented at the Interservice/Industry Training, Simulation, and Education Conference (I/ITSEC), Orlando, Florida.
Black, S., &amp; Darken, C. (2024). Scaling artificial intelligence for digital wargaming in support of decision-making. Naval Postgraduate School. Retrieved from arXiv:2402.06075
Chen, Y., &amp; Chu, S. (2024). Large language models in wargaming: Methodology, application, and robustness. Proceedings of the CVPR Workshop 2024.
Weller, D., Meltschack, M., &amp; Schwindling, D. (2024). Leveraging large language models for enhanced wargaming in multi-domain operations

NAML 2025

Agents for Navy Applications – A Practical Guide

hierarchical reinforcement learning (hrl)

cyber-attack mitigation

naval ai agents

### Welcome to the 10th annual workshop on Naval Applications of Machine Learning!

NAML is the premier annual event showcasing current machine learning research relevant to Naval applications. We welcome participation from the defense community, industry, and academia. The NAML workshop is an opportunity for researchers from Navy research and development centers and our partner organizations to exchange technical ideas and to come together as a team. 
Naval Information Warfare Center Pacific is partnering with AFCEA San Diego chapter to host the workshop.
#### On-Demand Access

* This is an on-demand site and all the recorded content from the workshop is available here. 
* Access to the recordings is exclusively available to registered participants.
* Each registrant has received an email notification, with instructions on how to access the content.

[![](https://assets.underline.io/markdown_image/1/image/60cd12aa376c191d3467ea2ee60944ea.png)](https://naml2026.org/)

The goals of NAML are:

* To spread awareness of current machine learning research relevant to Naval applications
* To connect machine learning researchers with experts in military requirements
* To build and strengthen collaborations within the defense research community

#### *"What we do as a Navy is a human business. Don’t pass up an opportunity to come together as a team. Don’t pass up leadership opportunities, and don’t outsource them to a PowerPoint. If you have an opportunity to come together as a team, to increase belonging, do it."*
Admiral John M. Richardson, 31st Chief of Naval Operations

This is an on-demand site. All recorded content from NAML 2026 is available here and access to it is exclusively available to registered participants.

**Please log in using the email address you registered with.** If you have any issues logging in, please contact us at <naml2026@underline.io>.

Let’s celebrate the past decade by investing in the next. 🚀

Please log in to explore this event.

This is an on-demand site. All recorded content from NAML 2026 is available here and access to it is exclusively available to registered participants.

**Explore the latest advancements in machine learning tailored to Naval applications 🔎**

With on-demand access to NAML 2026, you can watch all lectures & browse poster presentations at your convenience, while also engaging directly with presenters via the Underline platform - ask questions, exchange ideas, and stay connected.

**If you still haven’t registered, you can do so now.**
👇
**Take advantage of a special one-time offer: full access for just $99, with complimentary registration for students, government, and military employees.**

**Get access here: https://naml2026-ondemand.paperform.co/**

Please check your spam folder, or contact our support team at **<naml2026@underline.io>** if you registered, but still haven't received an access email.

Let’s celebrate the past decade by investing in the next. 🚀

Please register!

NAML 2026

The 10th Annual Workshop on Naval Applications of Machine Learning (NAML 2026) was held in San Diego, California, March 2-5, 2026. NAML is a forum for interdisciplinary experts from government, industry, and academia to discuss results of ongoing research on theory and applications of machine learning to address some of the most challenging problems faced by our naval forces. 



Agents for Navy Applications – A Practical Guide
Mark Tabladillo, Ph.D.
Cloud Solution Architect, Microsoft
Support Technical Advisor, Flank Speed
Abstract
This presentation explores the application of artificial intelligence (AI) agents in enhancing naval operations through the integration of Autogen and Azure AI Foundry. With the increasing complexity of naval missions and the rising threat of cyber-attacks, these tools provide a scalable and efficient solution for real-time decision-making and operational resilience.
The session highlights how hierarchical reinforcement learning (HRL) frameworks and modular agent architectures can be used to address the multi-level challenges of naval coordination. Autogen’s capabilities for creating and managing AI agents are combined with Azure AI Foundry’s robust infrastructure to demonstrate practical applications in ship coordination and cyber-attack mitigation.
A live demo showcases a simulated naval scenario where a fleet must navigate through a high-threat zone while responding to cyber-attacks. The demonstration illustrates how individual ship agents implement defensive measures, a command agent oversees fleet coordination, and large language models (LLMs) analyze intercepted communications for situational insights. These interactions highlight the agility and adaptability of AI systems in dynamic operational contexts.
The presentation concludes with a discussion on the scalability and future potential of Autogen and Azure AI Foundry in multi-domain naval operations, emphasizing their role in maintaining decision-making superiority and mission continuity in complex environments.
Citations:
Black, S. E. (2024). Mastering the digital art of war: Developing intelligent combat simulation agents for wargaming using hierarchical reinforcement learning (Doctoral dissertation). Naval Postgraduate School. Retrieved from https://hdl.handle.net/10945/73072
Black, S., & Darken, C. (2023). Scaling intelligent agents in combat simulations for wargaming. Paper presented at the Interservice/Industry Training, Simulation, and Education Conference (I/ITSEC), Orlando, Florida.
Black, S., & Darken, C. (2024). Scaling artificial intelligence for digital wargaming in support of decision-making. Naval Postgraduate School. Retrieved from arXiv:2402.06075
Chen, Y., & Chu, S. (2024). Large language models in wargaming: Methodology, application, and robustness. Proceedings of the CVPR Workshop 2024.
Weller, D., Meltschack, M., & Schwindling, D. (2024). Leveraging large language models for enhanced wargaming in multi-domain operations

technical paper

### Welcome to the 9th annual workshop on Naval Applications of Machine Learning!

NAML is the premier annual event showcasing current machine learning research relevant to Naval applications. We welcome participation from the defense community, industry, and academia. The NAML workshop is an opportunity for researchers from Navy research and development centers and our partner organizations to exchange technical ideas and to come together as a team. 
 Naval Information Warfare Center Pacific is partnering with AFCEA San Diego chapter to host the workshop.

The goals of NAML are:

* To spread awareness of current machine learning research relevant to Naval applications
* To connect machine learning researchers with experts in military requirements
* To build and strengthen collaborations within the defense research community

#### *"What we do as a Navy is a human business. Don’t pass up an opportunity to come together as a team. Don’t pass up leadership opportunities, and don’t outsource them to a PowerPoint. If you have an opportunity to come together as a team, to increase belonging, do it."*
Admiral John M. Richardson, 31st Chief of Naval Operations

Please log in to explore this event. You need to log in with the email address you registered with. Access credentials have been sent to your email. 

Please be sure to check your spam and other email folders if you do not see an email confirmation right away.

Registration for this workshop has now closed. We appreciate your interest and hope to welcome you at a future event. If you’d like to hear about similar upcoming events, please check our website or join our mailing list.

Registration for this workshop has now closed.

The 9th annual workshop on Naval Applications of Machine Learning is an opportunity for researchers from Navy research and development centers and our partner organizations to exchange technical ideas and to come together as a team. The goals of NAML are to spread awareness of current machine learning research relevant to Naval applications, to connect machine learning researchers with experts in military requirements and to build and strengthen collaborations within the defense research community.

TorchSig is a radio frequency (RF) machine learning (ML) software library that performs dataset generation and model training for signal detection and modulation recognition. This tutorial will be in two parts, beginning with a TorchSig overview followed by live demonstrations of the software. The TorchSig overview will include a brief history, future plans and then cover details such as datasets, signal generation, transforms and models. The overview leads into the demonstration portion which will walk through a series of publicly available Jupyter notebooks. The tutorial will show how to generate both wideband and narrowband datasets, how to train YOLO models with TorchSig data and how to use the pre-trained models.

TorchSig: Open-source Software for RF Machine Learning

MAGTF (Marine Air Ground Task Force) is a mission where different Program of Records collaborate. The MCTSSA/MCSC Test & Evaluation team creates test plans from the Requirement PDF documents. NPS-CS team used Computer Science, AI/ML concepts to explore automation and analysis to assist in the human intensive process.

System Requirements Clustering with Machine Learning and Architecture Design

Our presentation will describe and demonstrate the OLIVE platform and its machine learning capabilities. OLIVE is a Government off the Shelf (GOTS)/Unlimited Government Rights multi-media processing system in wide use in the DoD and Intel Community. OLIVE is a multi-media software development and deployment platform designed to be 1) readily deployable and transition technology from basic research to software customers can use, 2) flexible and easy to integrate into a wide range of customer systems, 3) modular, with the ability to update core components separately from the base system/platform, 4) efficient, facilitate the reuse of capabilities with a shared architecture for audio, text, imagery and video, 5) inexpensive, such that new innovations can be built off the existing base easy as possible so that the technology we’ve already developed is free and remains USG Unlimited Rights, 6) supported by a team of experienced researchers, software engineers and quality assurance personnel to provide training, documentation, maintenance and facilitate implementation and testing of new features. 
OLIVE is a transition platform, and as technology matures in our core research programs, we port the successful and robust components as “plugins” within the OLIVE framework to make them readily accessible to new and existing USG end-users and mission sites. Plugins include approximately thirty different capabilities, including voice biometrics (speaker recognition), language recognition, translation, transcription, face detection and recognition, topic detection and others. We have integrated technology from over twenty distinct projects from DARPA, IARPA, SOCOM, FBI, IC, DoD, etc. OLIVE runs natively on Windows, Android and Linux operating systems.
OLIVE is a software server using TCP/IP-based API with various client interfaces. OLIVE has a high-level API for both Java and Python and a lower-level API using Google ProtoBuffers and ZMQ that are compatible with almost all modern programming languages. The OLIVE framework is composed of several key parts: 1) Idento, SRI’s Core Deep Learning/AI and signal processing library, 2) the core OLIVE Server, which connects plug-ins with end user systems through an API. The enterprise API exposes capabilities to end-user interfaces and the core services API enable developers to integrate new capabilities into OLIVE., 3) Plugins, the independent modules that define recipes for specific capabilities (e.g. language ID). Each plugin has domains, that target a specific set of conditions or language. Each end user may use different plugins and there are often mission-specific plugins or domains. 4) GUIs, OLIVE has four GUIs developed by SRI and is integrated into many other end user GUIs. The OLIVE GUIs may be used directly by customers, though many use the OLIVE APIs directly to expose plug-ins to end users.
OLIVE is powered by multiple artificial intelligence models capable of running off-line in USG systems. Deployments have ranged from enterprise cloud scenarios for the IC to Android deployments for special operations. We will provide details on our approaches to deploying high-end, and high-accuracy AI capabilities in reduced resource compute environments and customizing models and algorithms to the military environment with high-noise acoustics, uncooperative targets and specialized language use.

OLIVE: a government-rights multi-media processing platform for the deployment of Artificial Intelligence

Tracking user activity requires generating precise logs of each interaction a user has with a machine including initiating background processes, recorded events from an input device, or recording within a graphical user interface. Some activity logging tools offer a compact storage format and tabulated data structure, but the information within these logs can still be complex. Oftentimes logs generated by interactions with graphical user interfaces are organized hierarchically within the tabulated columns. To improve the accessibility and comprehension of information within these datasets, a standard approach to parsing is introduced that translates the raw format into uniformly structured nested JSON objects. This approach involves sanitizing the data and then recursively applying a structure translation to each level within the hierarchy until every tier has been successfully converted.

We explore collecting user activity data to infer cybersecurity workflows and demonstrate domain knowledge can be elicited from subject matter experts’ (SME) user activity to train an autonomous assistant capable of providing decision support relating to the workflow. To provide additional context when training this model, a programmatic technique for generating natural language summaries of the parsed datasets is also proposed. The implementation offers a context-rich overview of the logged user activity by combining the previously generated, easily accessible JSON objects with predefined templates. Efforts were also made to support quality management of the generated summaries. A filter was developed for extracting the most relevant records or omitting irrelevant ones from parsed datasets before summarizing them. Cleaning was also introduced for aggregating similar records into one or omitting duplicates. Lastly, a concatenation functionality allows for record grouping so that summaries can be generated in varying user defined lengths.

This context-rich user activity data was used to generate an interactive timeline to showcase a SME’s workflow and allow them to label its components. To support end users in understanding the distribution of labeled workflow components, supplementary visualizations, such as bar graphs and heat maps, were also generated for visualizing the frequency of these components. Two different density metrics were used to define the frequency: a count of how often a particular label, or workflow component, was mentioned within all labels or a count of the number of records a label encompasses within its corresponding dataset.

To conclude, tracking human-computer interactions results in complex datasets that can be utilized to support classification of SME workflows. Such workflows offer potential for automated decision assistance by simplifying the creation of user activity summaries, quality management methods, and a variety of visualizations. In the future, efforts would be aimed at providing more detail within the generated summaries and improving functionality of the user interface for the interactive timeline.

Informing Cybersecurity Decision-Making: Parsing, Summarizing, and Visualizing User Activity Data

INFAMI is an intelligence generation framework that automates certain processes to help analysts and operators prioritize relevant targets and improve decision-making timelines. 
* ATR Management: Managing & deploying a library of ATRs and analytics
* Fusing/correlating tracks from multiple INT sources including GEOINT, SIGINT, and AIS/ADS-B/Thresher
* Facilitating tasking air and space assets and sensors via M2M interfaces; automated cross-cue and target custody
* Prioritizing targets to defeat adversary CCD
* Leveraging existing, mature interfaces and messaging systems (including OMS/UCI) 
* Rapidly integrating new ATRs, analytics, and existing capabilities into the framework

INFAMI: Accelerating TCPED on the Edge

Understanding how Large Language Models (LLMs) store and retrieve knowledge is key to improving their transparency and reliability. This work explores four interpretability techniques: Attention Visualization, Probing Classifiers, Activation Patching, and Causal Tracing. These methods reveal how LLMs track dependencies, encode linguistic and factual knowledge, and store information across layers. The findings show that knowledge is distributed but can be localized, though key questions remain about how it is structured and disentangled. This research provides insights into model explainability and highlights areas for future study.

Interpretability of Large Language Models: Frameworks and Experimental Insights

Active sonar systems, composed of transmit and receive sonobuoys or other platforms, have been developed to provide effective anti-submarine warfare (ASW) performance against small, quiet, red-force submarines in the harsh clutter-saturated littoral and the deeper open ocean. The studies of sonar field performance are largely based on "acoustic" performance predictions via sonar equation modeling, when in fact, there are also "tactical" considerations which, though ignored, are sure to have significant impact on blue-and red-force mission effectiveness. Tactical behaviors for sonar operations need to be learned, understood and incorporated into sonar field performance studies. To address this need, this research develops a reinforcement learning (RL) algorithm to represent AI red-force submarine evasive behavior. The RL algorithm aims to determine a red-force submarine trajectory towards its goal while minimizing the detection vulnerability of the submarine in a monostatic sonar field in a CZ environment. While this research applies a modern reinforcement learning technique to model the submarine evasive behavior in a specific scenario, more complex and diverse scenarios need to be studied to fully understand the capabilities and limitations of the approach. Different assumptions on blue-force sensors being localizable by the red-force submarine will be tested in the performance evaluation.

RL Based Red-Force Submarine Trajectory Optimization in a Monostatic Sonar Field 

The pursuit to generate visually pleasing images that are good representatives of a given category has been one of the driving forces of developments in generative vision models. One such model that is based on Generative Adversarial Networks (GAN) is called the Auxiliary Classifier Generative Adversarial Networks (AC-GAN). AC-GAN uses an auxiliary classifier to enable sampling from a class conditional distribution and is known to generate high-quality aesthetically pleasing synthetic data. Due to its simplicity, it is one of the popular GAN architectures for generating data. However, it has been shown that AC-GANs tend to learn data distributions that are biased away from the decision boundary of the auxiliary classifier. This bias away from the decision boundary is the unwanted by-product of a loss term in the AC-GAN formulation that favors high image quality and visual class coherence. That is, the AC-GAN generator is rewarded for generating samples that are towards the center of the decision region and discouraged from sampling near the class boundaries defined by the auxiliary classifier. In this work, we propose using a boundary aware loss in conjunction with the AC-GAN losses to mitigate the bias. We show that the introduction of the boundary-aware loss into AC-GAN formulation provides better sampling coverage within the underlying data distribution, increasing data diversity and reducing bias

A Boundary Aware Loss for Mitigating the Boundary Bias in AC-GAN

How organizations can manage the transformation driven by technological change.

Innovation and Organizational Transformation

The DoD Digital OnRamp Platform is a cutting-edge solution, designed to connect innovative private sector organizations to DOD opportunities, while also making private sector capabilities discoverable by Department of Defense (DOD) users.  The Platform will bridge the knowledge gap and provide a demand signal from the DOD to commercial/academic technology communities.  Leveraging the latest advancements in AI technologies (AI/ML/LLM/etc.), this platform aims to simplify and enhance the process of matching private sector capabilities with DoD opportunities, and vice versa.
 
The DOD Digital OnRamp Platform will provide an AI-enabled, secure, scalable, and efficient environment to facilitate:
•	Private Sector Matching to DOD Opportunities: Utilizing multiple, publicly-available data sources such as FPDS.gov, SBIR.gov, etc., the platform will provide users with: 1) matches  to relevant opportunities and technologies, 2) links to online training to help the user respond to the specific matched opportunities and 3) links to physical government offices within 100 miles of their address that can provide additional assistance, such as a Naval Tech Bridge, Air Force Spark Cell, DOD Innovation OnRamp Hub, APEX Accelerator, Small Business Administration office, etc.  
•	DOD Mission-Related Technology Queries:  Utilizing DOD and commercial datasets, the platform will provide DOD users with responses to their queries regarding technology, products and capabilities available from commercial companies and academia.  
•	Optimized Engagement Pathways: The platform will identify pathways to engage further for DOD users, i.e., the availability on certain DOD contracts or commercial options, as well as a collaboration layer to communicate with private sector organizations which have an account on the platform. 


Premium content

Downloads

Next from NAML 2025

TorchSig: Open-source Software for RF Machine Learning

Similar lecture

Rapid Prototyping for Naval AI Wargaming: Multi-Agent Scenario Learning and Counterfactual Reasoning with Python and AutoGen