AI platform for defense intelligence and tactical decision-making
Vannevar operates an agentic AI platform purpose-built for defense and national security missions, trained on a petabyte of adversarial data. The tech stack—Elasticsearch, PyTorch, TensorFlow, Hugging Face, vLLM, and AWS—reflects a mature ML infrastructure focused on real-time inference and retrieval at scale. Active hiring tilts heavily toward senior engineering (11 roles, mostly senior/lead level), signaling aggressive scaling of deployments into government authorized systems while wrestling with compliance hardening, retrieval latency, and the operational complexity of embedded tactical software.
Notable leadership hires: Growth Director, Growth Lead
Vannevar builds agentic AI software for U.S. national security and defense operations. The platform delivers targeting support, mission planning acceleration, and homeland and Indo-Pacific defense capabilities, currently deployed across more than 125 government and national security installations. The product surface spans information retrieval (Elasticsearch, OpenSearch), ML inference (PyTorch, TensorFlow, Hugging Face), and tactical decision support. Core operational challenges center on scaling deployments into compliant government cloud environments, optimizing search and embedding latency, and embedding software into existing defense programs and workflows.
Vannevar's stack includes PyTorch, TensorFlow, Hugging Face transformers, vLLM for inference, and ONNX for model portability. The platform runs on AWS infrastructure with Elasticsearch and OpenSearch for retrieval.
Vannevar supports more than 125 national security and U.S. government deployments. Current project focus includes scaling deployments to government authorized systems and expanding pilot engagements into larger programs.
Other companies in the same industry, closest in size