HEMIS AI / MISSION AUTONOMY FOR UNMANNED TEAMS

Machines will fight in teams.
Someone has to teach them how.

We build the intelligence that lets unmanned teams carry out a mission on their own, across air, ground, and sea, link or no link. The machines decide. A human stays in command.

Manifesto

By 2030, machines will fight the war. The side with superior intelligence and superior coordination wins.

01 / The reality

One operator per aircraft does not scale. Machines have to execute the mission on their own.

You cannot beat a hundred cheap drones with one operator flying one drone. The best aircraft in the world is still one aircraft and one target. Only a team beats numbers.

The asymmetry does not come from money. It comes from who is putting lives on the line. Machines are attritable. They have no emotion and nothing to lose.

A human approves every strike. Our adversaries may not show the same restraint. So the AI we build with restraint has to win at the front.

Our adversaries will buy this or build it. It is a matter of time. They will not give us time to build it after.

In Ukraine, engagement success went from 10 to 20 percent to 70 to 80 percent once terminal guidance stopped depending on the operator and the link.1

02 / The thesis

India is building platforms first. That is right. The industry needs it.

Platform makers are solving one hard problem. Intelligence is another one. Nobody solves both as a side project.

AI-enabled is not good enough. Good enough gets beaten 90 percent of the time. You can pay any price for an exquisite system. If it does not deliver, it is worthless.

You cannot field a research kit the day you need it. The intelligence has to be trained on real missions, on Indian terrain, with Indian tactics, today.

AI is as good as the data it trains on. Our borders, our terrain, and our adversaries are ours. We cannot outsource that data.

03 / What we build

Nightrun runs unmanned teams. Air, ground, and sea, fighting as one unit. A UGV on the ground and UAS overhead in different roles, one team.

Coldridge is where Nightrun trains. A simulation platform for training and tactical rehearsal in the new regime of war, so that nothing in the field is a surprise.

UAS is the starting platform. Heterogeneous unmanned teams are the point.

We build this for the Indian Army. It reaches them on the platforms they already buy.

04 / Platform makers

You build the platform. We build the intelligence that runs it as part of a team. No redesign.

  1. CSIS, March 2025. Forbes and the Kyiv Independent note the gain is largest for novice pilots.

ROADMAP

From a controlled area to a team the Army can field and send into battle.

  1. TODAY

    In the air.

    A team of UAS on a mesh, on one mission: cooperative search and detection of a defined target, in a controlled area. Demo video coming soon. Coordination policies train daily in our own simulation environment against opposing tactics, thousands of runs a week.

  2. PHASE 1

    Partner platforms.

    Nightrun runs on partner platforms over their existing command link, with no airframe change. A published interface, so any maker can bring a platform to the team. The first mixed team: aircraft from different makers on one mission.

  3. PHASE 2

    Field trials.

    Unmanned teams on real terrain, in real weather, against jamming, with the operators who will use them. Operators run their own tactics against the team. What breaks goes back into training.

  4. PHASE 3

    Ground and sea.

    Unmanned ground and surface vehicles join the same team. No mission is fought in one domain, so autonomy cannot stop at one. One team across air, ground, and sea, under one commander.