Real problems.Work I’m proud of.

The agent talks. The whole deal has to stay in sync.

A real-estate agent can describe a deal in a few sentences. Carrying that conversation through a transaction is a much bigger engineering problem. The property, people, money, documents and next action all have to agree—even as terms change and the agent on the other side gets involved.

I’m Varea’s CTO. I built the engineering team and lead the technical direction, with hands-on responsibility across the AI, backend and product experience. Together, we took an initial voice prototype into a transaction platform with agent workspaces, connected document workflows and the operational tools behind them.

Making AI work inside the deal.

One answer can contain several facts. A correction can affect the contract, an addendum and what the agent sees next. Property records, contacts and saved preferences can help fill the gaps, but they must not casually overwrite what the agent just said.

That is where much of my engineering work sits: extracting meaning from conversation, handling corrections and related-document data, and connecting voice actions to saved product state. I work on the transitions too—starting or resuming a deal, moving between documents, reading back terms and handing work over for review or signature.

Several people. Several documents. One transaction.

The platform’s workflow work spans purchase contracts, buyer–broker agreements, listings, clauses and addenda, counteroffers, inspection responses, disclosures and pre-closing walkthroughs. Those journeys have different participants, stages and signing requirements.

We built the agent portal and a scoped workspace for the counterparty agent, alongside document editing, PDF previews and signature integration. The challenge is carrying the right context into each action: which document is being changed, who may change it, which terms are current and what the next person needs to do.

The tools behind the product matter, too.

We also built an admin studio for configuring document templates, fields, prompts, formulas, conditional rules and signer information. That gives the team a way to manage the document system as the product grows.

Call histories, traces and extraction replay make AI behaviour inspectable. The team can investigate what was said, what the system extracted and where a call went wrong. I also turned the product’s complexity into a developer handoff covering workflows, architecture, responsibilities and acceptance scenarios.

My contribution went into the patent application, too.

I wrote the technical overview and worked directly with patent counsel on the architecture description and diagrams. I reviewed the drafts in detail, clarified how the system works, and corrected the technical account so it reflected what we built.

A US provisional patent application covering the system was filed on 9 March 2026. I’m proud that my contribution reached from designing and building the technology to explaining it for the application.

I stay involved in delivery, demonstrations and agent onboarding. I’m proud of the team we built and the system we put around the idea. My responsibility reaches from the conversation an agent has to the software, documents and people that have to act on it.

The product. The architecture. The next decision.

Conductr helps renewable-energy projects move through a complicated network of developers, experts, documents and approvals. The founder needs a platform that reflects that business—and someone who can decide how it should evolve. That is my role as fractional CTO.

Decide what is worth building.

I work directly with the founder on priorities and product direction. I challenge requests, propose alternatives and decide with the team what to build now, what to make configurable and what can wait. A customer asking for a new workflow does not automatically need a separate product. We work through the business case before turning it into engineering work.

Give the business room to evolve.

I led the move beyond the earlier no-code platform and the architectural work around it. We mapped the business rules behind organisations, permissions, subscriptions, files and workflows, then translated them into a plan the engineering team could execute. The responsibility was preserving how the business worked while giving us more control over its development.

Our delivery connects project requests, partner offers, shared datarooms and commercial workflows. Customer releases introduced co-location and project-sale requests, partner NDA requirements and notifications when shared documents change. I stay close to the details that make those journeys work together.

Look after the resources, too.

I also secured AWS Activate cloud credits for Conductr, with approval confirmed in October 2026. Finding support for the infrastructure belongs alongside choosing the architecture: both affect what the company can afford to build next.

Stay close when customers arrive.

I work with the founder on onboarding, configuration and delivery priorities as customers put the platform to use. My job is to connect what the business needs, what the software should do and what the team can deliver. The founder has a technical partner in those decisions, from product strategy through implementation.

I built the company that built the products.

A founder can see the business opportunity and still need someone to build everything around it: the first product, the architecture, the team and a realistic plan. As co-founder and CTO of CodeCooperation, I took responsibility for that whole picture.

From zero to an engineering organisation.

I built the technical organisation from the ground up. During my time there, it grew to 31 engineers across six product teams, with our work spanning more than 20 startups and products. I recruited and mentored engineers, set architecture and review standards, and stayed hands-on with implementation.

We also built an engineering operation across Germany and Tunisia. Establishing that team meant recruitment, working practices and collaboration across borders. Tech216 and the GIZ-supported Invest for Jobs initiative documented that part of the company’s growth.

A partner to the founder before the first line of code.

I worked with founders on product scope, technical choices, budgets and roadmaps: what the first version needed to prove, what could wait, and which people would make it happen. I also mentored founders through Founder Institute. Helping someone make a sound decision—and then putting a team behind it—is a large part of my work.

Products that made it into the world.

Seniorenplatzfinder helped connect families seeking care with care providers. Ilioa brought music streaming to businesses. Translatly connected clinicians with interpreters through video. These became working products and released applications, each built around a different business and a different set of users.

Translatly reached a pilot at Goethe University Hospital in Frankfurt. I’m a named co-author of the peer-reviewed study published in JMIR Formative Research in 2025, which credits my software architecture and development work.

I’m proud of the products—and of building an organisation capable of delivering them. When a founder needs technical leadership, I bring experience of building both the software and the company behind it.

The whole operation was the brief.

A property team has buildings to manage, owners to answer and properties to market. Its day can disappear into finding documents, sorting emails and carrying information between systems. Across Verwaltungspartner Nord and Zukunftsverwalter, I became the technology partner responsible for turning those recurring problems into useful software.

An internal GPT that knows where to look.

I built an internal AI assistant around property information, uploaded documents and connected business systems. Staff can ask questions, retrieve supporting material and see the sources behind an answer. The work includes document handling, spreadsheet questions, persistent conversations and integrations with Impower and Casavi. It gives the team a way into the information spread across its operation.

From incoming messages to work someone can act on.

I built email classification inside Outlook around three practical questions: which department, how urgent, and whose responsibility? The routing reflects the business, from accounting enquiries to property-management requests.

Alongside that, I delivered AI telephone intake and a staff dashboard connecting call summaries, property and contact matching, callbacks and Casavi tickets. I carried it through to client handover, with a guide for the team’s daily review and follow-up. Capturing a request and making it usable belong to the same job.

Property photos in. Marketing material out.

I also built and deployed Property Studio: photos, plans and a plain-language brief become a proposed storyline, prepared images and a generated film. The team can review the sequence, revise the material and approve generation. That puts a repeatable marketing workflow around the AI, with the property’s own source material at its centre.

The systems underneath are my responsibility, too.

The partnership extends into system migrations, property data and document transfers, and integrations with existing software. That includes making Microsoft email work inside Immoware and verifying incoming mail, outgoing delivery and sent copies.

This is the value of having me close to the business: I can spot the opportunity, build the tool, connect it to the operation and stay involved as people use it. The result is a set of capabilities the team can work with across its day.

Expertise was the idea. A working marketplace was the job.

Skillbid gives people a direct way to learn from experienced professionals: ask a question, offer a price and receive expertise through text, video or a live conversation. Making that exchange work means connecting the experience on both sides with a business the founder can operate.

As fractional CTO, I took responsibility for product engineering across expert onboarding, discovery, paid questions, recorded and live sessions, calendars, payments, receipts and administration. My work connected the customer experience to the systems needed to run the marketplace.

Make the business model work in the software.

I worked through how money should move with the service: reserving payment, collecting it at the right stage, handling exceptions and giving the mentor and operator usable transaction records. I carried that work into implementation and verification. The payment, the answer and the person delivering it all have to belong to the same transaction.

Give expertise another way to reach people.

I also built mentor-specific AI chat around the expert’s own material, connected to public profiles, a visitor trial and conversation history. People can begin exploring the knowledge behind a profile, alongside the platform’s human services. That required the knowledge retrieval and the product experience around it.

Own the decisions behind the build.

I advise the founder on product priorities, architecture and where engineering effort belongs. When a major rewrite was discussed, I recommended clarifying the product goals and focusing on attracting users before committing to more development. We worked through team continuity and the consequences of changing technology. Protecting the business’s momentum is part of technical leadership.

A product out in the world.

Skillbid reached public beta, and MTH Accelerator welcomed the company into its programme. Founder Uwe Flade publicly acknowledged my engineering contribution at launch. I carried product decisions into working software and continue to provide the technical leadership behind the next stage of the business.

The doctor has already said it. The team shouldn’t have to start again.

In a specialist practice, the work continues after the conversation. Dictation becomes a letter. A letter needs review. The phone brings an appointment request, a prescription request or a callback. Information passes through several people before the next step is done.

For Praxis Ballo & Böck, I built two products around those handoffs: Medikta for medical documentation and Mira for telephone intake. I worked directly with the doctors and practice team on the product, engineering and daily workflow.

From spoken notes to a usable clinical document.

Medikta turns German medical dictation into structured, editable letters. I built the surrounding workflow: configurable sections, reusable text blocks, saved drafts, clinician approval and an assistant interface for review and export. The doctor can dictate and correct; the practice team receives a document it can work with.

Clinicians used Medikta regularly and helped shape it. Specialist vocabulary, preserving the doctor’s wording and keeping corrections intact mattered. I also worked with the practice’s IT provider on the audio connection. My responsibility followed the product into the working environment.

Give reception a structured handoff.

Alongside Medikta, I developed Mira to capture appointment, prescription and other telephone requests into records for staff follow-up. I built the conversation flow and dashboard around what the team needs to see, review and act on.

Doctors and medical assistants tested it with complex requests and medication names. Their feedback shaped the conversation and the handling of unfinished work. Mira’s work is a tested telephone-intake product; Medikta is the product clinicians used in their documentation workflow. Clinical decisions and approval stay with the practice.

Built around the people doing the work.

The achievement is the connection between speech, documents and the next person’s task. I took responsibility for those handoffs with the clinicians themselves. My earlier collaboration with Dr. Ballo also includes Translatly, a separate healthcare product that reached a university-hospital pilot and a published study I co-authored.

They needed a department. I took the lead.

FairPlay Digital brought me in to take over software and AI and build the way that part of the company operates. The assignment reaches across the business: processes, client offers, team readiness and technical delivery. A company needs to know what it can sell, who can deliver it and how the work will be run.

Build the process behind the promise.

I connect the stages that can otherwise fall between commercial and technical teams: understanding the client’s problem, assessing feasibility, defining scope, planning the work and reviewing delivery. I establish responsibilities and technical review practices, and push for company-controlled repositories so the business retains control of its software.

Put engineering judgment into the offer.

I shape software and AI proposals with a clear first step, an implementation approach and a scope the team can deliver. I separate discovery, a pilot and a larger platform so the company can make a credible commitment and the client understands what they are buying.

For an enterprise knowledge-assistant opportunity, that meant working through SharePoint integration, existing access permissions, source documents and how to evaluate answers. The offer has to account for the real environment it will work in.

Get the team ready for the work.

I work through the skills and people each project needs, how responsibilities should be divided and whether the team is equipped for the proposed scope. Staffing, architecture and delivery planning belong in the same conversation. My role is to bring those decisions together before the company commits to a build.

Stay close enough to deliver.

I remain hands-on in implementation where needed. Work on the training storefront was one example within a much wider remit. The larger responsibility is leading the department: shaping what FairPlay can offer, organising the people and processes behind it, and carrying that through to delivery.