Linchpin Tactical AI

Latent AI shall integrate technologies to optimize and secure machine learning (ML) runtimes on edge hardware. The effort will better refine the traceability, observability, replaceability, and automated consumption (TORC) requirements for Project Linchpin. The scope includes the integration of the Latent AI Efficient Inference Platform (LEIP) into the Government IL5/IL6 system, design/development of ML algorithms for Army mission-relevant use-cases such as Automated Target Recognition (ATR), optimization and securing the model runtime, and benchmarking results on target edge hardware. The project will characterize and define the AI bill of material (AI-BOM) and standardize the methods to benchmark ML algorithms on edge hardware. The project will also demonstrate the use of LEIP in a portable, ruggedized, deployable kit (Ruggedized AI Kit), to fine-tune and deploy ML models in the field. The final deliverable will be a MLOPs workflow for Linchpin, with tools defined to train, deploy, optimize, and deploy AI models.