# Andreas Merentitis — person profile

**Research snapshot:** 24 August 2026  
**Publicly associated organization:** OLX Group  
**Publicly listed location:** Berlin, Germany  
**Role reported in public sources:** VP Data / Chief Data Officer Europe (CDO Europe), OLX Group  
**Identity note:** The email address supplied for this research is `andreas.merentitis@olx.com`; this profile does not independently verify ownership of that address. The findings below are based on public professional pages and publications, not private correspondence or private social-media data.

## Executive readout

Andreas Merentitis presents publicly as a technically deep data and AI leader who is increasingly focused on turning machine-learning capability into measurable marketplace and customer outcomes. His profile combines a long research and engineering background with senior operating responsibility at OLX. The strongest recurring themes are proprietary, domain-specific data; experimentation and deployment speed; user-centered product design; AI agents; personalization; fraud prevention; and building organizational capability rather than merely shipping features.

The clearest evidence of his current thinking comes from his public LinkedIn posts and a March 2026 interview with Romanian technology publication start-up.ro. His public voice is energetic and optimistic, but not purely hype-driven: he repeatedly contrasts impressive demos with business value, user trust, reduced friction, and measurable outcomes. His writing is usually accessible to a broad business-and-technology audience, even when the underlying subject is advanced machine learning.

## Tone & persona

**Observed tone:** optimistic, pragmatic, technically literate, and strongly impact-oriented. He writes with the confidence of an executive who has spent years in applied data science, but often explains ideas through everyday user experiences: searching for a home, posting a car listing, finding a job, or avoiding fraud. That makes the voice more human and product-oriented than purely infrastructure- or research-oriented.

He is enthusiastic about AI and the future of marketplaces, using celebratory language such as “proudest moments,” “exciting time in Tech,” “real customer value,” and “we're creating it.” At the same time, he explicitly rejects AI for its own sake. In a public post, he says, “The technology itself is never the story. The impact is,” and urges readers to “stop counting features, start counting impact” [[1]](https://www.linkedin.com/posts/andreas-merentitis-37b262b_how-do-i-know-when-ai-is-actually-working-activity-7485564598436331520-WQSg). This is a useful summary of his persona: ambitious about technology, skeptical of vanity metrics, and interested in evidence.

He also comes across as collaborative and ecosystem-minded. His posts praise teams, invite discussion, highlight colleagues and external speakers, and frame industry events as opportunities to share what is “actually working,” rather than as one-way self-promotion. The emotional energy is high, but the intended posture is generally inclusive: explain the idea, show the use case, credit the team, then invite others to connect or compare experiences.

**Inference:** The public record supports describing him as an authoritative but approachable technical executive. “Authoritative” is supported by his senior role, research record, and repeated use of operational metrics; “approachable” is an inference from his conversational openings, emojis, praise of collaborators, and open questions at the end of posts.

## Vocabulary

Characteristic or recurring terms in his public communication include:

- **AI / GenAI / machine learning / data intelligence / agentic AI**
- **marketplaces**, **customer value**, **user value**, **customer journeys**
- **proprietary data**, **vertical data**, **data loop**, **local market liquidity**, **pricing history**, **transaction patterns**, and **conversion signals**
- **intent, context, and nuance**
- **product–market fit at scale**, **experiment**, **production**, **roll out**, **ship faster**, and **execution speed**
- **optimize, scale, disrupt** — the three-pillar framework he used to describe OLX’s AI innovation approach [[2]](https://www.linkedin.com/posts/andreas-merentitis-37b262b_ai-datascience-innovation-activity-7433047469963411456-aGXX)
- **meaningful connections**, **remove friction**, **time given back**, **trust**, and **impact**
- Product and technical names used in his OLX communication include **CompassGPT**, **AutoIQ**, **AutoGPT**, **Ad2Video**, **Video2Ad**, **Jobs Matchmaking**, **Talk2data**, **Toqan**, and **FrAIday**.

He favors verbs that imply movement and delivery: **build, launch, deploy, optimize, scale, simplify, remove, eliminate, measure, learn, experiment, create, and connect**. His language often pairs a technical mechanism with a human consequence—for example, natural-language search with “less confusion,” or AI-generated listing content with time returned to customers and families.

## Structure

His social posts commonly follow a recognizable pattern:

1. **Open with a question, contrast, or vivid user scenario.** Examples include “How do I know when AI is actually working?” and “What if finding a home felt less like navigating a database and more like having a conversation?” [[1]](https://www.linkedin.com/posts/andreas-merentitis-37b262b_how-do-i-know-when-ai-is-actually-working-activity-7485564598436331520-WQSg) [[3]](https://www.linkedin.com/posts/andreas-merentitis-37b262b_aiinnovation-realestate-marketplacetransformation-activity-7462812611466854401-6niU)
2. **State a thesis in plain language.** The thesis is usually that AI should solve a real customer or organizational problem, not simply add a feature.
3. **Give concrete examples and metrics.** He names products, markets, use cases, adoption, speed, acceptance, retention, or quality measures. Metrics in his posts should be treated as company- or speaker-reported claims rather than independently audited results.
4. **Introduce a framework or causal model.** The “Optimize / Scale / Disrupt” framework and the “capture proprietary signals → build vertical models → act on predictions → measure outcomes → reinforce and improve” loop are representative examples [[2]](https://www.linkedin.com/posts/andreas-merentitis-37b262b_ai-datascience-innovation-activity-7433047469963411456-aGXX).
5. **Close with a forward-looking position and an invitation.** He often ends by saying the future is being created now, then asks readers what they are seeing or invites them to connect.

His older LinkedIn articles are more editorial and explanatory. The 2020 article “We know what you might buy tomorrow: here’s why” begins from the desired user experience—relevant, customized items with less hassle—and connects that experience to predictive personalization [[4]](https://www.linkedin.com/pulse/we-know-what-you-might-buy-tomorrow-heres-why-andreas-merentitis). His 2021 “OLX Data Science blog posts” article acts as a curated gateway to engineering topics including a jobs recommender, Item2Vec recommendations, and explainable AI [[5]](https://www.linkedin.com/pulse/olx-data-science-blog-posts-andreas-merentitis).

## Background

Public LinkedIn profile data reports that Merentitis has been at OLX since 2017, progressing from **Head of Applied Data Science, OLX Group Berlin Hub** (November 2017–January 2019), to **Director of Data Science** (January 2019–March 2020), and then **VP Data / Chief Data Officer Europe** from March 2020 onward [[6]](https://de.linkedin.com/in/andreas-merentitis-37b262b). The profile identifies Berlin as his base and OLX Group as his current employer.

Before OLX, the same public profile reports roles as **Senior Data Scientist and Senior Research Scientist at Zalando SE** (September 2015–November 2017) and **Senior Data Scientist / Senior Researcher at AGT International** (December 2011–August 2015). Earlier work included a network and systems administrator role at the Hellenic Army Information Support Center, adjunct teaching at the Technological Educational Institute of Piraeus, and senior research at the University of Athens [[6]](https://de.linkedin.com/in/andreas-merentitis-37b262b).

His education is rooted in the National and Kapodistrian University of Athens. Public profile data reports a Bachelor’s degree in Computer Science (1999–2003), a Master’s in Computer Engineering (2003–2005), and a PhD in Computer Engineering (2005–2010), with honours reported for the graduate degrees. The profile’s education description identifies a thesis on reliability optimization of wireless sensor networks under high-performance and low-power requirements [[6]](https://de.linkedin.com/in/andreas-merentitis-37b262b).

His research record spans dependable and energy-efficient embedded systems, remote sensing and data fusion, neural networks, and reinforcement learning. Verified arXiv records list him as a co-author of **Stochastic Maximum Likelihood Optimization via Hypernetworks** (2017/2018) and **A Bandit Framework for Optimal Selection of Reinforcement Learning Agents** (2019) [[7]](https://arxiv.org/abs/1712.01141) [[8]](https://arxiv.org/abs/1902.03657). The public profile also reports Senior Member status with IEEE and the IEEE Signal Processing Society and a first-place result in the 2013 IEEE GRSS Data Fusion Contest [[6]](https://de.linkedin.com/in/andreas-merentitis-37b262b). Its structured metadata additionally lists a first-place Kaggle result; I found that claim on the profile but did not independently validate it in a primary Kaggle record, so it should be treated as **profile-reported**.

## Communication style

His current LinkedIn writing is formatted for fast scanning: short paragraphs, rhetorical questions, bold conceptual contrasts, arrows, checkmarks, hashtags, and occasional emojis such as 🤔, 🚀, 🤖, 🐙, and 💬. Posts are often medium-to-long rather than terse, but the prose is broken into compact units and lists of examples. He uses headings inside posts—“Three Pillars,” “Real Examples, Real Impact,” and “What Makes This Possible”—to make a technical narrative legible.

The style shifts by audience. In executive and product-oriented posts he explains concepts through marketplace use cases and outcomes. In research and engineering contexts he uses specialized vocabulary such as hypernetworks, reinforcement-learning agents, recommender systems, embeddings, and explainable AI. His 2018 LinkedIn article about Search2Vec and a PyData meetup demonstrates a community-and-knowledge-sharing mode, while the 2021 article directs readers to OLX engineering material about production data science [[5]](https://www.linkedin.com/pulse/olx-data-science-blog-posts-andreas-merentitis) [[9]](https://www.linkedin.com/pulse/search2vec-olx-group-pydata-meetup-berlin-andreas-merentitis).

There is no strong evidence in the sources reviewed of a heavily formal, legalistic, or highly academic public voice. Nor did this research find a verified public X/Twitter account attributable to him. LinkedIn is the clearest public channel.

## Key interests & expertise

- **Applied AI for marketplaces:** search, recommendation, personalization, content generation, fraud prevention, lead intelligence, and customer-service or shopping assistants.
- **Proprietary and localized data:** Merentitis argues that marketplace advantage comes from data others do not possess—local price histories, liquidity, transaction behavior, dealer signals, and conversion patterns—not simply from access to the same foundation models.
- **Agentic and conversational experiences:** His public examples move marketplaces away from rigid filters and forms toward natural-language search, conversational buying, voice agents, and task-oriented assistants.
- **Product thinking for data scientists:** In the March 2026 start-up.ro interview, he says senior data scientists need to understand UX, domain problems, and project-management concerns, and increasingly need to interpret data creatively [[10]](https://start-up.ro/omul-care-are-cultura-generala-viitorul-in-domeniul-it-si-in-tehnologie-hi-andreas-merentitis-olx-group-hi/).
- **Organizational AI adoption:** His FrAIday post frames AI enablement as a learning problem. He favors hands-on experimentation, distributed learning, internal “AI Professors,” and cross-functional participation over training that is only lecture- or completion-based [[11]](https://www.linkedin.com/posts/andreas-merentitis-37b262b_fraiday-ai-fraiday-activity-7469733865394282496-BPqE).
- **Core technical foundations:** machine learning, algorithm design and optimization, ensemble learning, remote-sensing data fusion, anomaly detection, distributed systems, neural-network optimization, and reinforcement learning.

## Notable quotes or positions

The following are public statements attributed directly to Merentitis; numbers in the newer posts are self-reported OLX/company figures and are not independently audited here.

- **“The technology itself is never the story. The impact is.”** This captures his outcome-first stance [[1]](https://www.linkedin.com/posts/andreas-merentitis-37b262b_how-do-i-know-when-ai-is-actually-working-activity-7485564598436331520-WQSg).
- **“Stop counting features, start counting impact.”** He uses this as a proposed standard for evaluating AI products [[1]](https://www.linkedin.com/posts/andreas-merentitis-37b262b_how-do-i-know-when-ai-is-actually-working-activity-7485564598436331520-WQSg).
- **“The future of marketplace search isn't about better filters. It's about no filters at all.”** This expresses his preference for conversational, intent-aware product experiences [[3]](https://www.linkedin.com/posts/andreas-merentitis-37b262b_aiinnovation-realestate-marketplacetransformation-activity-7462812611466854401-6niU).
- **“Agential AI does not replace people; it strengthens their decisions.”** This is an English translation of a statement quoted in the Polish Business Insider report about his CLAIM AI remarks; the source’s wording and translation should be checked before using it as a verbatim English quotation [[12]](https://businessinsider.com.pl/technologie/nowe-technologie/claim-ai-w-lizbonie-jak-autonomiczna-sztuczna-inteligencja-zmienia-sposob-w-jaki/687j5h0).
- In the start-up.ro interview, he describes AI as a new stage of technological development rather than a total reinvention, with the major difference being **scale** and the ability to extract insight from data across sources and structures. He also argues that technical and humanistic/product disciplines need to “meet in the middle” so teams can build experiences users actually want [[10]](https://start-up.ro/omul-care-are-cultura-generala-viitorul-in-domeniul-it-si-in-tehnologie-hi-andreas-merentitis-olx-group-hi/).

## Social presence

**LinkedIn:** The verified public profile is [linkedin.com/in/andreas-merentitis-37b262b](https://de.linkedin.com/in/andreas-merentitis-37b262b). Its public page exposes a substantial set of authored articles and posts. Older authored pieces include “PyData meetup @ OLX tech hub Berlin” (2017), “Han presenting @ OLX Group” (2018), “Search2Vec at OLX Group – PyData Meetup Berlin” (2018), “PyData Berlin meet-up hosted by OLX” (2019), “Speaking at D3” (2019), “We know what you might buy tomorrow: here’s why” (2020), and “OLX Data Science blog posts” (2021) [[6]](https://de.linkedin.com/in/andreas-merentitis-37b262b). Public article metadata on the page showed 5,475 followers at the time of retrieval; this is a time-sensitive platform count, not a durable biographical fact.

**Recent public activity:** 2025–2026 posts emphasize AI impact, AI Agents Week, Toqan, conversational marketplace search, AutoGPT, CompassGPT, AutoIQ, FrAIday, and CLAIM AI. He describes a keynote at Tech Spirit Barcelona and promotes the recording of the event sessions [[13]](https://www.linkedin.com/posts/andreas-merentitis-37b262b_agenticcommerce-olxinnovation-ai-activity-7430162736787066881-i-qi). He also promoted CLAIM AI 2026, an invite-only Lisbon summit about AI and marketplaces [[14]](https://www.linkedin.com/posts/andreas-merentitis-37b262b_claimai2026-agenticcommerce-olxinnovation-activity-7423984735124606976-Vb00).

**Talks, interviews, and media:** Corinium listed him as a CDAO Frankfurt 2024 speaker on real-world GenAI applications and challenges, with a focus on customer experience and business innovation [[15]](https://www.coriniumintelligence.com/content/unlocking-cdao-frankfurt-power-data). The March 2026 start-up.ro interview is the richest media source found for his views on the changing role of data scientists, polymathic skills, AI product design, and OLX use cases [[10]](https://start-up.ro/omul-care-are-cultura-generala-viitorul-in-domeniul-it-si-in-tehnologie-hi-andreas-merentitis-olx-group-hi/). Research-oriented visibility comes through the arXiv papers and IEEE publications; public search did not surface a verified personal blog, podcast archive, or separate X/Twitter presence.

## Sources and confidence notes

Primary or near-primary sources used: public LinkedIn profile and authored LinkedIn posts/articles [[6]](https://de.linkedin.com/in/andreas-merentitis-37b262b), [[1]](https://www.linkedin.com/posts/andreas-merentitis-37b262b_how-do-i-know-when-ai-is-actually-working-activity-7485564598436331520-WQSg), [[2]](https://www.linkedin.com/posts/andreas-merentitis-37b262b_ai-datascience-innovation-activity-7433047469963411456-aGXX), [[3]](https://www.linkedin.com/posts/andreas-merentitis-37b262b_aiinnovation-realestate-marketplacetransformation-activity-7462812611466854401-6niU), [[4]](https://www.linkedin.com/pulse/we-know-what-you-might-buy-tomorrow-heres-why-andreas-merentitis), [[5]](https://www.linkedin.com/pulse/olx-data-science-blog-posts-andreas-merentitis), [[9]](https://www.linkedin.com/pulse/search2vec-olx-group-pydata-meetup-berlin-andreas-merentitis), [[11]](https://www.linkedin.com/posts/andreas-merentitis-37b262b_fraiday-ai-fraiday-activity-7469733865394282496-BPqE), [[13]](https://www.linkedin.com/posts/andreas-merentitis-37b262b_agenticcommerce-olxinnovation-ai-activity-7430162736787066881-i-qi), and [[14]](https://www.linkedin.com/posts/andreas-merentitis-37b262b_claimai2026-agenticcommerce-olxinnovation-activity-7423984735124606976-Vb00). Academic sources: arXiv records [[7]](https://arxiv.org/abs/1712.01141) and [[8]](https://arxiv.org/abs/1902.03657). Independent reporting and event context: start-up.ro [[10]](https://start-up.ro/omul-care-are-cultura-generala-viitorul-in-domeniul-it-si-in-tehnologie-hi-andreas-merentitis-olx-group-hi/), Business Insider Poland [[12]](https://businessinsider.com.pl/technologie/nowe-technologie/claim-ai-w-lizbonie-jak-autonomiczna-sztuczna-inteligencja-zmienia-sposob-w-jaki/687j5h0), and Corinium [[15]](https://www.coriniumintelligence.com/content/unlocking-cdao-frankfurt-power-data).

Confidence is **high** for the LinkedIn-listed employer, role family, location, authored content, public quotes, and academic co-authorships because those claims were visible in public profile pages or primary publication records. Confidence is **moderate** for exact current title wording and current activity metrics because professional profiles and social counts change. Confidence is **lower** for claims sourced only from profile metadata, such as the Kaggle result, and for any personality judgment: tone and persona are informed inferences from public communication, not private knowledge.