Wild Card

Dismounted infantry and SOF units routinely interact with international partners in scenarios where linguists are unavailable and the units rely on 1-2 bilingual soldiers to operate. During combat operations, this lack of language capability is a failure point that limits true interoperability. Current translation solutions are not suitable for use while units are maneuvering, assaulting an objective, or reacting to contingencies. I am searching for a DOD Sponsor to enable  our Hacking for Defense team to explore a solution from January to March 2026. 

When small, dismounted infantry or special operations units integrate with international partners, language gaps at the edge slow maneuver and break interoperability. Interpreters are often unavailable, out of position, or too few, and relying on one or two bilingual soldiers creates single points of failure. In critical moments like first contact or a casualty event, communication devolves to “pointy talky,” increasing risk to mission and force.

There are numerous commercial solutions that tackle real time translation for the military however they generally fall into two buckets:

1. some sort of enterprise level or networked radio over IP (RoIP) solution;  or
2. a tactical variant of a Neural Machine Translation app/ translation LLM on a smart phone requiring cloud access and sometimes running on device.

These are not the right solution for a dispersed infantry platoon conducting a forward passage of lines through an international partner during large-scale combat operations, or for an ODA directing a partner force while reacting to an ambush during a UW campaign. 

Through a cross-functional team of a machine learning engineer, and an embedded systems engineer, and myself (a Green Beret Veteran & JD/MBA student at the University of Chicago), via our Hacking for Defense (H4D) course, we will build a prototype small, kit-worn offline hardware device that takes audio in and delivers the target language out through input- and output-agnostic paths that would prevent an infantryman wearing gloves from fumbling with a smart phone to communicate with a partner while maneuvering in a rainy jungle at night

Our Course End Goals: 

• Conduct 30 or more interviews with military and industry experts to map constraints and define success criteria.
• Identify a solution to fit DOTMLPF and existing unit tactics, techniques, and procedures.
• Build and test a prototype device.
• Create a roadmap to pursue non-dilutive SBIR/STTR funding and a CRADA for MVP development and testing. 

Prototype Key Performance Indicators

• Sub 10% Word Error Rate
• Sub 2 second latency
• 12-hour battery life
• Sub 3lbs weight

1. Latency under 2 seconds on low SWaP compute while keeping battery life and thermals in check
2. Maintaining accurate (90% accuracy) audio output from processing audio in austere conditions (gunfire, vehicular noise, stressed screaming individuals
3. Maintaining low weight in a form factor that operators actually would be willing to wear on their kit.
4. Balancing 1-3 while meeting AOR language and dialect requirements
5. Handling jargon, call signs, slang, borrowed words, and military specific terminology in multiple languages
6. Creating a simply user interface for hands free or minimal hands use in an austere environment at night.
7. Integration with military audio outputs (PELTOR/ Ops-Core Headset)
8. Balancing low price with US or friend-shored supply chain
9. Ruggedization to MIL-STD-810 standards