AOTEAROA NEW ZEALAND · DUAL-USE SENSING PLATFORM

Built to listen
anywhere it matters.

HARK started with a question: what if we never stopped listening to the forest? The same edge-AI platform that tells a kiwi call from a stoat's footstep, on-device, on solar, with no signal — is now being asked to listen for something else entirely. One sensor. Two frontiers.

100+
Species classified on-device
10 yrs
Battery life, LiFePO₄
0 bytes
Raw audio leaving the field
2
Divisions, one core
PLATFORM

ONE PLATFORM

Every division runs on the same nervous system.

Wildlife and Sentinel are two products with one shared spine: a field-hardened edge-AI listening device that classifies sound where it happens and only ever ships home the results. Nothing below changes between a reserve in Fiordland and a perimeter fence — only what the model is trained to hear.

Edge AI

On-device classification

Runs a Perch-class bioacoustic model locally across three ultrasonic mics (192kHz). No cloud round-trip, no dependency on signal to think.

Power

Solar + LiFePO₄

~116mW active draw, 10-year cell life. Deploy once, forget it's there — in bush or behind a fence line.

Comms

4G, mesh, and beyond

Results — not raw audio — sync when connectivity allows. Mesh relays extend coverage across terrain with no towers.

Hardening

Field & tamper resistant

Stainless enclosure rated for kea, rats, weather, and curious hands. Built for terrain nobody visits twice a year.

Data

Results-only by design

Classifications and metadata leave the device — not recordings. A privacy posture conservation needed first; security needs it too.

Updates

OTA firmware & models

New species, new signatures, new capability — pushed to fielded units without a truck roll.

DIVISION I — WILDLIFE

DIVISION I — CONSERVATION

HARK Wildlife

"Our forests are speaking.
Are you listening?"

The founding product, unchanged in purpose: continuous, autonomous acoustic monitoring for the places conservation teams can't visit every week. Where a manual bird count happens once or twice a year, HARK listens every night, all year, and tells the difference between a kiwi call and a predator's footstep the moment it happens.

  • Deployed with the Department of Conservation, Predator Free 2050 Ltd, University of Auckland, and Ngāti Maru — pairing mātauranga Māori with on-the-ground sensing.
  • 800 Trust, HARK's registered NZ charity, planted 28,000 trees in 2025 with Air New Zealand — the mission funds itself back into the whenua.
  • First devices reach conservation partners in 2026.
Use case

Invasive species alerting

Stoat, rat, and possum calls flagged in near-real time so trapping teams move before a nest is lost.

Use case

Population trend tracking

Year-round call density replaces once-a-year counts with an actual trend line for funders and iwi.

Use case

Remote reserve coverage

Solar autonomy puts sensors in terrain too remote or steep for regular foot patrol.

DIVISION II — SENTINEL

DIVISION II — SECURITY & DEFENCE

HARK Sentinel

"Our perimeters are speaking.
Are you listening?"

What tells a kiwi call from a stoat's footstep can just as easily tell a service vehicle from a snare being cut, or a drone rotor from wind through the canopy. Sentinel repackages the same hardened, solar, edge-AI core around a new set of listening tasks — for teams who also can't be everywhere, all the time.

  • Concept division, pre-pilot — positioned for agencies already comfortable with HARK's field-hardened, offline-first approach to sensing.
  • Same privacy posture, now a security feature: results-only sync means no raw audio trail to secure, subpoena, or leak.
  • Targeting pilot conversations with conservation-security and rural-policing partners across the Asia-Pacific from 2027.
Use case

Anti-poaching & safari reserves

Snare activity, off-track vehicles, and gunshot events flagged to ranger radios across reserves too large to patrol continuously.

Use case

Border & perimeter monitoring

Unattended sensors along remote border stretches classify foot and vehicle traffic where fences and cameras run out of power.

Use case

Rural policing

Gunshot and off-road vehicle detection for the vast rural ground urban acoustic-sensing networks were never built to cover.

Use case

Coastal & marine enforcement

A buoy-mounted variant listens for engine signatures around marine reserves — the same conservation logic, pointed at the water.

CONCEPT ROADMAP — NOT YET AVAILABLE

Capability under evaluation

R&D

Gunshot triangulation

Multi-sensor timing to localise a shot to sub-meter accuracy in under 100ms.

R&D

UAV acoustic signature ID

Classifying drone rotor signatures separately from wind, birds, and aircraft.

R&D

Vehicle fingerprinting

Engine and tyre acoustic profiles to distinguish patrol, service, and unauthorised vehicles.

R&D

Mil-spec mesh encryption

AES-256, frequency-hopping mesh for contested or connectivity-denied ground.

R&D

LEO satellite backhaul

An uplink option for theatres with no cellular or mesh path home at all.

R&D

Tamper self-wipe

On-capture, the unit purges local model weights and keys before it can be exploited.

ONE CORE, TWO OUTPUTS

SAME CORE, DIFFERENT QUESTION

One capability, read two ways.

Nothing about the underlying sensing changes between divisions — only the label the model is trained to apply, and who the alert goes to.

Core capabilityWildlife reads it asSentinel reads it as
Sound classificationSpecies identificationThreat & vehicle classification
Anomaly detectionPredator incursionIntruder or breach event
Trend analyticsPopulation change over timePattern-of-life & activity trends
Direction & triangulationLocating a call in dense bushLocating a shot, engine, or breach
Results-only syncLow-bandwidth conservation reportingDefensible, low-signature evidence trail
Solar autonomyUnmanned reserve coverageUnattended perimeter or border coverage
ROADMAP

WHAT WE'RE BUILDING

From one device to two divisions.

2026

HARK Wildlife devices ship

First units reach DOC, Predator Free 2050, and university research partners across Aotearoa.

2026 — H2

Platform hardening for dual-use

Mesh encryption, enclosure ruggedisation, and OTA model pipeline generalised beyond bioacoustic species sets.

2027

Sentinel pilot program

First fielded units with anti-poaching and rural-policing partners; gunshot and vehicle classification in limited pilot.

2028

Satellite backhaul & fusion sensor pod

LEO uplink option and a thermal-acoustic fusion variant for connectivity-denied, high-value sites.

INVESTORS

THE OPPORTUNITY

Two markets. One sensor network.

HARK is a hardware company with a data platform's economics. Splitting into two divisions doesn't split the R&D spend — it doubles the addressable ground the same core technology can be sold into.

Illustrative market sizing

Conservation & biodiversity tech
~US$3–4B
Perimeter & defence sensing
~US$9–11B
Combined addressable
~US$13–15B

Illustrative figures for discussion purposes only — not sourced market research. To be replaced with commissioned sizing ahead of raise.

Why this, why now

  • Edge-AI inference cost has dropped far enough to run classification on solar power in the field — the constraint used to be compute, now it's go-to-market.
  • Conservation funding rewards continuous monitoring data; defence and policing budgets are shifting toward low-cost distributed sensing over expensive fixed infrastructure.
  • One R&D spend, two procurement channels — grant and ESG capital for Wildlife, government and enterprise contracts for Sentinel.
  • A charitable arm (800 Trust) and an existing DOC/Predator Free 2050 track record de-risk the brand entering a second, higher-scrutiny market.
01

Hardware

Device sales into conservation agencies, reserves, and — pending pilots — security and defence procurement.

02

Platform

Subscription access to the results dashboard, analytics, and fleet management across both divisions.

03

Contracts

Higher-margin, multi-year government and enterprise agreements as Sentinel moves from pilot to deployment.

See the full model, the roadmap detail, and the raise.

Request the investor briefing