The genuinely AI-relevant piece of this defense contract is buried in the technical requirements: Cohere Technologies must deliver a "Layered Inference Sensing" capability that converts raw Delay-Doppler radar returns into real-time 3D target tracks with automatic classification and confidence scoring, an applied inference task layered on top of what is otherwise a conventional RF and telecom waveform contract. For practitioners tracking where classical signal processing and applied ML increasingly overlap, defense sensing contracts like this are a useful signal of where funding and technical requirements are heading, even though neither public source detailed the underlying model architecture.
What happened
Cohere Technologies, a wireless technology company known for OTFS (Orthogonal Time Frequency Space) modulation and unrelated to the AI startup of the same name, won a $28 million contract from the U.S. Department of War's FutureG Office, led by Tom Rondeau, according to Light Reading and TelecomsTechNews. The contract funds a multi-waveform Radio Access Network prototype, combining Cohere's Pulsone (Zak-OTFS) waveform with standard OFDM in a single baseband, to detect uncrewed aerial systems using existing commercial 5G/6G spectrum rather than dedicated radar hardware. The work builds on an earlier NSF VINES Phase 2 research program and must comply with the Pentagon's open RAN architecture requirements.
Technical context
The prototype uses an Integrated Sensing and Communications (ISAC) approach, in which the same RF signals used for communication double as a sensing channel. Per Light Reading's reporting, the contract calls for a Layered Inference Sensing system that processes raw Delay-Doppler radar data into real-time 3D tracking with automatic target classification and confidence scoring, an ML-adjacent inference task, though neither source disclosed model architecture, training data, or whether it relies on learned models versus classical detection algorithms.
For practitioners
This is primarily an RF and telecom infrastructure contract rather than an AI-native one, but it illustrates how defense sensing programs increasingly specify inference and classification requirements rather than just raw signal capture, which is where applied ML skills intersect with traditional radar and RF engineering. Teams working on signal classification, sensor fusion, or edge inference for aerospace and defense use cases may find the ISAC and Layered Inference Sensing approach worth tracking as a funded reference architecture.
What to watch
Field trial results from Cohere's planned Mobile Test Platform, any published technical detail on the inference or classification model used, and whether the FutureG office extends similar ISAC-based sensing contracts to other vendors.
Key Points
- 1Cohere Technologies, a wireless RF firm distinct from the AI company, won a $28 million DoD FutureG contract for a drone-detection RAN prototype.
- 2The system layers a machine-inference target classification module on top of a multi-waveform 5G/6G sensing and communications architecture.
- 3Defense sensing contracts increasingly bundle classification and inference requirements into RF hardware programs, a trend worth tracking for applied ML engineers.
Scoring Rationale
Primarily an RF/telecom defense contract, not an AI-native milestone; scored down from the initial 6.7 to reflect that the AI/ML angle (automatic inference-based target classification) is a modest technical component rather than the core of the story, while still crediting genuine applied-ML relevance for practitioners in signal classification and sensor fusion.
Sources
Public references used for this report.
Practice with real Ad Tech data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Ad Tech problems
