We are honored to share that the Department of Defense (DoD) Defense Innovation Unit (DIU) has awarded Fiddler with a Success Memo for completing the Automated Machine Learning for Mine Countermeasures Operations (AMMO) MLOps prototype with the U.S. Navy.
The DIU in collaboration with the Navy posted Machine Learning Operations (MLOps) as an area of interest on May 31, 2022 and selected Fiddler as one of the awardees of the prototype proposal.
The goal of the AMMO prototype is to build an MLOps pipeline toolset capability to rapidly retrain the Automatic Target Recognition (ATR) ML models powering the U.S. Navy’s mine countermeasures (MCM) application, thereby empowering the Navy to quickly adapt to changing undersea threats in diverse theaters.
As part of the implementation, Fiddler was initially installed on the security-enhanced AWS GovCloud environments and then successfully validated for all three phases of Government Acceptance Testing (GAT), including functionality confirmation, security posture, and deployment automation.
In addition, the Fiddler AI Observability platform for federal agencies was recently deemed Awardable by the DoD Chief Digital and Artificial Intelligence Office (CDAO)’s Tradewind Solutions Marketplace to accelerate the procurement and adoption of AI/ML, data, and analytics at the DoD.
The following key capabilities of Fiddler have been validated and are leveraged in the AMMO workflow:
- Image Explainability: Enterprise grade on-demand explainability to help model developers and in the future, mission operators, understand and trust the decisions of classification and object detection models
- Image Monitoring: Patented data drift monitoring of image embeddings integrated with visual debugging of these embeddings using UMAP. This helps identify operational changes in the AI model’s behavioral and perform root cause analysis of any issues quickly
These capabilities were also validated in a live demonstration of the AMMO prototype which showed a 97% decrease in the time needed to update the ATR models with this MLOps toolset — from six months to a few days.
The Success Memo followed by the production migration of Fiddler along with the remaining vendors in the AMMO MLOps toolset represents a substantial step forward in the Navy’s MCM capabilities in subsea seabed warfare. While AMMO shows significant benefits in this use case, the methodology and implementation is expected to seamlessly extend to other use cases, sensor types, and modalities across the DoD. This allows the DoD to confidently deploy ML and subsequently LLM models safely and reliably at scale. With AI policy top of mind following the Presidential Executive Order on Trustworthy Use of AI, Fiddler’s AI Observability platform also enables Federal teams to meet the outlined transparency mandates.
“IQT first invested in Fiddler in 2019 because we believed that explainability and transparency would be essential to the responsible adoption of Artificial Intelligence across the U.S. government. Fiddler’s recent success with Project AMMO is a clear demonstration of how AI can and should be built to support our national security in line with the President’s Executive Order on safe, secure, and trustworthy AI.”
— A.J. Bertone, Managing Partner, InQTel
"UUVs are critical to US defense strategy, particularly in the Indo-Pacific region, providing critical intelligence capabilities. Working with Fiddler AI and other partners, we've crafted an AI model development and deployment tech stack that enables operational effectiveness."
— Joel Meyer, President of Public Sector, Domino
“Project AMMO brings together Fiddler's Observability platform with Latent AI's Tactical Edge Optimization platform to enable more comprehensive end-to-end AI/ML solutions for the DoD.”
— Jags Kandasamy, CEO, Latent AI
Project AMMO highlights the DIU, and DoD’s commitment to bringing next generation AI and ML tooling to enable rapid deployment of responsible AI.
For more information about AI monitoring and explainability for Government, contact our AI experts.
![The AMMO and UGIS technology demonstration teams](https://cdn.prod.website-files.com/5e069c660d37a18c17207c29/66ad3309e545f09626c5f912_663bfe26636a27c0b19b2388_tQStnGwb4Etj11l05eue_dx00eZnbuACJFKGAeuKykanCCL6JQhLYldFux7Q8fxZJFBHNvhE4k97xQLS9OK8SeU0BYvNXhN1BNF4iZcZ8XpRahy7_jTRuWnkeU3lZJ48TeNL1yZahjXkedtg6_LFXPM.png)