Skip to content
shirman.dev
Building GeoNexo & HYJACK

Eitan Shirman · Product builder, Berlin

I turn complicated work into products people can actually use.

I design and build AI products, marketing systems and the interfaces that make them understandable. Right now that means GeoNexo and HYJACK.

See what I'm building Start a conversation
What is HYJACK?What is GeoNexo?Smaller experiments
Portrait of Eitan Shirman
Eitan ShirmanProduct direction, interface design and engineeringBerlin, Germany · Open to difficult product problems
GeoNexo · CTOHYJACK · BuildingLinkedIn ↗
Working acrossProduct · Design · Engineering
Building nowGeoNexo · HYJACK
Interested inWorkflows people route around

Selected work / 01

Two products.
One connected problem.

How do you turn noisy information into a decision someone can understand, edit and defend?

HYJACK · Marketing workflow product

Turn good source material into a campaign.

In development
Good references end up scattered across tabs, videos, screenshots and chats.

HYJACK brings references, transcripts and audience signals into one workflow, then helps a team shape the strategy, brief and creative without losing where the thinking came from.

Product designEvidence workflowsMarketing systems
What I do on this project

I am building the product, interface and connected workflow. A locally verified backend path now carries captured evidence into an approved plan, an editable brief and an unscheduled draft. The wider product experience is still being integrated.

  1. CaptureReferences, transcripts and screenshots
  2. IntelligenceSignals that keep their source attached
  3. Approved planA person signs off on the direction
  4. Editable briefStrategy you can change, not regenerate
  5. Unscheduled draftNothing publishes on its own

Local backend validationFixture-backed checks cover duplicate prevention, preserved edits, recovery and provenance. It is not production-validated yet.

GeoNexo · AI visibility productBuilding now

See how your brand shows up in AI answers.

GeoNexo helps teams understand when AI systems mention their brand, which sources shape those answers, and where important gaps remain.

What I do on this project

As CTO, I work across product direction, interface design, architecture and integration. The hard part is not collecting more data. It is turning a noisy field of answers, citations and competitors into a next step someone can understand.

Visit GeoNexo
Research · Neural encoding

Can brain-response research help creative teams ask better questions?

A non-commercial research and education project about neural encoding, media response and the limits of interpreting models such as TRIBE v2.

Research context only. No resale of model access, and no claim to predict individual behaviour or sales.

Independent work / 02

Useful experiments,
with honest labels.

Smaller products I built to answer a specific question. Prototype means prototype.

Local prototype

01 / Private knowledge

BrainDrop

A private knowledge system that captures messages, files, meetings and archives, then answers questions with links back to the original evidence.

Current state

The core provenance, review and retrieval system exists. The next honest test is simpler: use it every day for a week and prove that capture and retrieval stay reliable.

Personal prototype

02 / Commerce

Virtual try-on

A Chrome extension that replaces garment photos on a Vinted listing with generated try-on images, using one reference photo.

Current state

It detects listing images, runs edits one at a time, survives page rerenders, and can stop or restore the originals. It is a visual experiment, not a fit or sizing tool.

Website work

03 / Client delivery

Bauwucht

A clearer digital front door for a German flooring and renovation business, focused on its services, mobile experience and inquiry path.

Current state

Hands-on WordPress implementation and design refinement. The current work is about making the offer easier to understand and the mobile journey easier to use.

How I work / 03

Trace the evidence. Design the decision. Build the handoff.

  1. 01

    Find the real bottleneck

    Map what people do today, where context gets lost, and which decision the product must make easier.

  2. 02

    Make the system inspectable

    Show sources, states and tradeoffs clearly. People should understand what the software did before they trust it.

  3. 03

    Ship the smallest complete path

    Build one useful journey end to end, test the difficult edges, then expand from evidence instead of assumptions.

Have a difficult product or workflow?

Bring me the thing everyone currently works around.

Early products, tangled internal workflows, and systems that have become harder to use than the problem they were meant to solve.

Opens your email app, addressed to hello@shirman.dev. Prefer LinkedIn? Message me there ↗