I build operational systems that help organizations move faster. At Almaz Capital I managed deal-flow infrastructure for thousands of startups. Now I build AI automation workflows with n8n and Claude. Both come from the same instinct: find the friction, remove it.
I started as an industrial engineer, learning to treat every process as something measurable and improvable. That instinct followed me when I moved into venture capital operations at Almaz Capital in Berlin, where I spent four and a half years building the operational backbone behind a growing fund.
At Almaz I grew the startup pipeline from 300 to over 2,500 companies, automated roughly a third of our deal-flow operations, and built the data infrastructure that let the team make faster, better-informed investment decisions.
Right now I am combining that background with hands-on AI work: building n8n and Claude workflows that solve real operational problems. Early-stage and growth companies where operations is still being built from scratch are where I do my best work. The feedback loop is short, and a well-designed process can change how the whole team functions.
From the factory floor in the US to venture capital infrastructure in Berlin. Two very different environments, the same underlying drive to make complex systems work.
Personal automation projects and case studies from real operational work. Click any card to read the full case study. More to come as I take on pro-bono and freelance work.
Three linked n8n workflows: a daily job scraper (Arbeitnow + Ashby), an MCP server exposing the job board as five queryable tools, and a Chrome extension that captures any job posting with one click. Claude extracts metadata and saves the verbatim job text to Notion with full formatting preserved.
A full redesign of Almaz Capital's deal-flow infrastructure: data modelling, intake automation, enrichment pipelines, and reporting dashboards. The CRM grew 8x in four years without proportional growth in maintenance effort.
A ground-up redesign of receiving, tracking, and reconciliation at a Continental Tire facility. Lean Six Sigma tools surfaced the real error sources and a targeted fix achieved a sustained 40% reduction in discrepancies.
Starting to take on operational work with startups and SMEs: building automations, cleaning up pipelines, and streamlining ops. Case studies will be added here as engagements wrap up.
I am currently looking for full-time roles in operations, strategy, and AI-driven process work, with a focus on early-stage or growth-stage companies in Berlin or remote-friendly positions. I am also open to pro-bono and freelance projects.
Email is the best way to reach me. I typically respond within a couple of days. You can also find me on LinkedIn or on Xing for a fuller picture of my background.
The problem
Tracking a job search by hand means copy-pasting company, role, location, and salary from each posting into a spreadsheet. The same listing appears across multiple boards, so you end up doing it more than once. Low-value work that is easy to let slip.
What I built
Job Scraper. An n8n workflow that runs every morning at 7am. It pulls from Arbeitnow's public API and from Ashby pages for companies I track in a Notion CRM. A Code node filters the results to German roles and remote European roles. No AI involved; job type is derived from the source's own fields.
MCP Server. A separate n8n workflow that exposes the job board as five queryable tools: search by title, by description keyword, by date, get full details, list latest. This gives Claude and other AI agents direct read access to the board without writing custom API code each time.
Job Page Capture. A Chrome extension button that saves any job posting with one click. The extension parses the page into typed content blocks (headings, bullets, paragraphs) and sends them to an n8n webhook. Claude Haiku reads the blocks, extracts structured metadata (company, role, location, type, salary), and marks the exact block range where the real job content starts and ends. The job text is copied verbatim into a new Notion page, headings, bullets, and links intact. Claude identifies boundaries and pulls field values; it does not summarize or rewrite anything.
Duplicate detection checks company name, position title, and location against existing entries, with city name normalization to catch the same role listed under different spellings (Cologne vs Köln, Munich vs München, and so on).
Results
Stack
The situation
When I joined Almaz Capital, there was no standardized deal flow process. Deals moved through the pipeline without defined stages, no clear criteria for advancing them, and no assigned ownership at each step. Communication with founders happened ad hoc.
What I did
I built the deal flow process from scratch. That meant defining pipeline phases, setting the requirements for moving a deal from one stage to the next, deciding what actions had to happen at each step, and establishing who owned each deal at any given point. The team adopted it quickly because it gave everyone a consistent structure to follow rather than improvising for each deal.
Once the process was in place, it became clear how much execution time was going into repetitive manual work: entering companies into the CRM, enriching their data, setting reminders to revisit deals, and keeping founder communication timely across every pipeline stage from first contact to deal execution. I redesigned those workflows using CRM-native automation features, communication sequencing tools, and AI-native workflow automation (n8n, Zapier).
The largest gap I addressed was outbound. There was no structured outreach engine before. I built one: sourcing criteria, communication sequences for each pipeline stage, and the tooling to run it consistently at scale.
Results
Stack
The context
Over three years at Continental Tire I worked across Production, Purchasing, Engineering, and Corporate on a range of process improvement, systems, and planning initiatives. The role covered everything from hands-on Lean and Six Sigma projects on the production floor to building reporting infrastructure and leading multi-year investment planning.
What I did
I ran Lean and Six Sigma projects to reduce production losses and redesign workflows, applying root-cause analysis and training manufacturing teams in continuous improvement methods. I also served as SAP EWM Key User, diagnosing recurring system errors and documenting the corrective requirements that brought them under control.
On the planning side I built monthly and annual demand and capacity forecasting models, and Power BI dashboards that gave the team a real-time view of production KPIs. Previously the team ran reviews off weekly Excel files; the dashboards shifted decisions to daily and made bottlenecks visible before they became problems.
I also led the 5-year departmental investment planning process, coordinating with corporate committees across Quality, Capacity, and Maintenance to develop and get buy-in for multi-year capex priorities.
Results
Tools & methods
