Discuss your requirements
Sensor fusion / counter-droneConcept view

From scattered sensors
to one picture
that decides.

Tiktaalik AI builds the decision layer for counter-drone operations. It brings the sensors a site already runs into one current picture, supports the operator's judgement, and keeps a record of every decision. Operators stay in command; AI is optional.

Discuss your requirementsScroll to the picture ↓

First conversation, no deck required. Sensitive material moves to a secure channel.

01 / The problem

Each new sensor adds a screen, not a picture

01

More consoles, not more picture

Each new sensor adds another screen, another login and one more place to miss the contact that matters.

02

One object, many contacts

When three radars see one drone, the operator sees three contacts. Birds and false detections add more.

03

History that cannot be shown

After an incident, the team must show what was seen, how fresh it was, which rule applied and who had authority. A log of the last action cannot answer that.

A counter-drone system on a tripod beside a frozen shoreline, with a drone in the sky.
Concept diagram: scattered sensor observations with duplicate contacts resolving into one continuous track.
Observation / track

Many contacts.
One track.

Observations arrive from different sensors at different times. Duplicate reports converge on one object, and the track shows which sensors contributed.

Before
12 contacts
Fused
4 tracks
Decided
1 picture
Advice stays advice.Authority stays human.
02 / What the platform does

What the platform does

These are the outcomes a site gets. Together they give the operator one picture to act on.

01

One current picture

A live view of the airspace that shows how recent each contribution is, so stale data looks stale instead of current.

02

One object, one track

Duplicate reports from different sensors become a single track, with the sensors behind it visible.

03

Classes you can act on

Each track gets a class and a confidence level: drone, vehicle, bird or false detection. The platform says when it is unsure.

04

Prediction and urgency

Short-term behaviour and an urgency estimate help an operator judge which contact matters first.

05

Plan before you commit

A digital twin of the site calculates what a sensor set can see and lets a team rehearse layouts before buying hardware.

06

Advice with evidence

Recommendations arrive with their evidence and alternatives, labelled as advice rather than as an instruction.

03 / Human control

Nothing acts until four checks agree

A command passes only when four independent checks all say yes. Each answers a different question, and each can stop the command.

Command path / interactiveCOMMAND PASSES ALL FOUR
A check can stop a command at any pointAll four checks agree. A command may proceed.

Any check can stop a command. When the conditions cannot be confirmed, the platform refuses and records the refusal.

Next step

If the picture is spread across four consoles, we should talk.

Discuss your requirements

No obligation. A human reply, not a sequence.

04 / Evidence chain

Six statements that stay separate

Observation, belief, judgement, advice, command and outcome are different statements. Keeping them apart is what makes a decision reviewable.

A recommendation is never a command. Completed means the equipment reported completion; it does not mean the effect occurred.

  1. 01

    Observation

    What a sensor reported.

  2. 02

    Track

    The current picture of one object, including how fresh it is.

  3. 03

    Assessment

    A judgement about class, behaviour or urgency, with a confidence level.

  4. 04

    Recommendation

    Advice to a person: what, why, the evidence and the alternatives. Never a command.

  5. 05

    Command

    An action authorized by a person after all four checks.

  6. 06

    Outcome

    What happened, and separately what was observed to happen.

Concept diagram: a site boundary, sensor positions and overlapping coverage circles, with a proposed additional sensor.
Site model / sensor coverageProposals require acceptance
05 / Digital twin

Plan the site with evidence

Model the site before hardware is bought, so sensor placement is based on evidence rather than opinion.

  1. See coverage first

    Model the site and calculate what a sensor layout can see before any hardware is bought or moved.

  2. Compare options

    The same layout gives the same result, so options can be compared fairly.

  3. Accept or reject proposals

    The twin can suggest adding, removing or replacing sensors. A person accepts or rejects it, and the decision is recorded.

Planning only. Nothing changes until a person accepts a proposal.

06 / Test, never guess

Rehearse before anything goes live

A test environment lets a team exercise a deployment before it is connected to live sensors.

MODETESTEFFECTORS BLOCKED / SWITCH RECORDED

  1. Test without live traffic

    A test site accepts synthetic sensor traffic, so the team can practise without touching live data.

  2. A clear TEST state

    The screen shows TEST, never LIVE, and simulated data cannot enter the live picture.

  3. Measured results

    Results are checked against known truth: correct merges, missed objects and wrong classes.

Rehearsal is kept separate from live operations.

07 / Deployment

One design, many deployments

The platform fits the setup a site already has, from a single site to a group of sites that share a picture.

one site / local operation

A complete local picture. It keeps working without a network.

How it runs

Standalone

A complete local picture for a site that does not run a command-and-control system.

Integrated

Adds the decision layer beneath the command-and-control system a site already operates. That system keeps authority.

Topology and host

Deployment is adapted to the site's network and hardware, and is designed to run without a cloud connection.

08 / Delivery

Software, training and support

The platform is software. It is delivered as a package an installation team can run in the field.

01

Software package

Installer, integration support and updates for the site's environment.

02

Training

Hands-on training for the team that installs and supports the system in the field.

03

Support

Remote support inside the customer's security rules.

Tiktaalik AI supplies the software and training; installation and hardware integration sit with the installing partner.

A radar operator in a ship operations room, with screen data reflected across their face.
01Sense
02Track
03Fuse and classify
04Decide
05Act
09 / Questions

Questions we hear before the first meeting

Q01Does Tiktaalik AI replace our command-and-control system?

No. The platform runs standalone or alongside the command-and-control system you already operate. In an integrated deployment it adds tracking, fusion, classification and decision support to the data that system already receives.

Q02Which sensors and response systems does it work with?

The platform is designed to connect to the sensors and systems a site already operates. Every integration is scoped with the customer.

Q03Where does the human sit in the loop?

In control. By default a person authorizes any action before it is sent. A customer can define where automation is allowed, within its rules, and every change is recorded.

Q04Do we need AI to use the platform?

No. AI is optional and the platform keeps working without it. When AI is used, it runs on the customer's infrastructure rather than a cloud service.

Q05What happens when the network is cut?

Each site keeps working locally. Remote data is marked as stale, and actions that cannot be confirmed are refused and recorded.

Q06How do we start?

Start with a conversation about the site and the decisions it needs to support. A first deployment can be small, and capability can grow as the need is proven.

Start

Tell us about the airspace you protect

Describe the sensors you run, the system you answer to, and the decisions you need to support. We will reply with a scoped next step.

Discuss your requirements

No obligation. The first conversation is just a conversation.