From language to performance
Learned English at 16 and started translating voluntarily, just to keep using it. By the time I got to university I had one goal: build my own translation agency. What I didn't expect was the communication side of the degree. It got me hooked on how people process a message and what makes it land.
Turkish Airlines came next. A couple of months in, I noticed a pattern that didn't add up. A structural error in the ticketing system was silently firing false cancellation notifications at scale. I just can't ignore mismatching patterns. I reported it, got ignored, gathered more evidence, reported again. Someone finally picked the report up and I was moved to a newly created data analysis team. Lesson I kept with me: if the data says something is broken, you don't wait for permission to flag it.
Then I built the agency I had been aiming for. MM Translation: 120+ languages, 150+ freelancers, clients across legal, medical, technical and academic fields. Project coordination, quality control, dozens of simultaneous workflows across language pairs, all of it ran through me. When the agency needed clients, I stopped outsourcing the marketing and did it myself. That one decision is why I'm here.
Digital marketing turned out to be applied psychology with a feedback loop that tells you where people drop and why. After years of studying how people process information, that language came naturally. A year of campaign and strategy work later, the 13x ROAS I pulled on EN Tercume confirmed what I already knew. This is the work I want to keep doing.
AI sharpened the edge. I spent a year training large language models and evaluating their outputs, building multi-step prompt flows and assessing accuracy, truthfulness and hallucination inside RLHF and NLP pipelines. That gave me a practical understanding of these tools most marketers don't have. I engineer prompts, build agentic workflows and plug these tools into real marketing work. I also picked up cloud computing and built my own self-hosted n8n automation server in three days, running custom pipelines that replaced a stack of subscriptions and fed offline conversions back into GA4 and Google Ads.
Now in Berlin finishing my MSc in Digital Marketing at BSBI/UCA, pulling it all together: communication, data, consumer behaviour and the systems that interpret it. My thesis sits right on that intersection. AI-localized vs human-localized advertising, how automated and human translation shift ad performance across cultures. ELM, TAM and Hofstede frameworks, automotive brands as the test case.
The mindset behind the results
I don't start with tools. I start with thinking. Before anything gets built or launched, I sit with the problem until I understand what's going on under the surface. The back and forth, questioning my own assumptions, is where most of the work happens. Execution afterwards is the easy part.
When I take something on, I take ownership of the outcome, not just the tasks in front of me. If something adjacent is broken and it's dragging results, I fix it or flag it. At Turkish Airlines I reported a system error nobody asked me to find. On EN Tercume I built the website, the tracking, the SEO and the campaigns, because the project needed all of it. I'd rather have too many responsibilities than too few.
Data is the starting point for every decision. I measure first, develop options from what the numbers say, then act. When the Upkeystore data said the US market was bleeding budget, I built four simulation models before recommending a pause. Gut feeling is fine for generating hypotheses. Data confirms or kills them.
If a skill or a tool will make me more efficient, I learn it and put it to work. Fast. Claude Code, Cursor, GA4 audience setups, offline conversion pipelines, a self-hosted n8n automation server built in three days, none of these were things I knew before I needed them. The gap between never touched this and it's live and working is usually days, not months.
AI isn't an add-on in my workflow. It's the default. I use it to research, analyse, build and iterate at a pace that wasn't possible two years ago. The thinking is still mine, AI just speeds up the parts where speed wins.
Everything I ship gets sharper the next time. I keep looking for what could be cleaner, even when the first version worked. That's why my campaigns keep getting better, not just running.
What I work with
Ads & Analytics
MarTech & Automation
AI & Dev
Web & Design
Workspace
How I connect channels and data
Strategy & Growth
Every campaign starts with one number I need to move. ROAS, CPA, leads per week, whichever the business lives on. From there I reverse-engineer the budget, the funnel stages and the test priorities. Gut feeling generates the hypothesis, data decides what survives.
Paid Media & Social
Google Ads across Search, Shopping, PMax and Display, Meta Ads for top and mid funnel. I build campaigns from a clean keyword tree, negatives from day one, UTM tagging wired into GA4 from the first click. Boring structural work on day one is what makes scaling painless on day ninety.
Analytics & Tracking
Most marketers stop caring about tracking the moment the tag fires. I don't. GA4 events, audiences, custom funnels, GTM setup, Looker dashboards and offline conversion pipelines, these are what turn ad spend into a feedback loop. If the data isn't trustworthy, nothing downstream is either.
MarTech & Automation
This is where AI earns its budget. I built my own self-hosted n8n server to run offline conversion uploads, lead enrichment and content pipelines without a per-step subscription stack. Make and Zapier still show up when the ROI says so. HubSpot and Salesforce live on the CRM side. Every step a machine can do is one less hour I waste.
Web & SEO
I build fast static sites, wire tracking in properly and keep SEO sane. This portfolio is static HTML/CSS/JS/PHP on Cloudflare Pages, migrated from WordPress in a single day with Claude Code and Cursor. On client work I add Rank Math, schema and the research stack when the project needs them.
Leadership & Management
At MM Translation I ran a network of 150+ freelancers across 120+ languages, legal, medical, technical and academic clients running at the same time. Project coordination, quality control and dozens of parallel workflows all passed through me. When something dragged results I went after the system, not just the symptom. The bigger the scope, the better.
AI & Emerging Tech
A year as an AI Trainer and Prompt Engineer on RLHF and NLP pipelines gave me a practical view of what these models are good at. I engineer prompts, build agentic workflows and plug them into real marketing work. My MSc thesis sits on that same edge, AI-localized vs human-localized advertising and how each shifts ad performance across cultures.