T-Digital by Deutsche Telekom · official title: Senior DevOps Engineer — Cloud Infrastructure
This is a public version of my CV. For the latest role-tailored résumé, get in touch
Senior Cloud Platform & Security Engineer with 12+ years of experience, focused on the platform side of AI: running RAG and LLM workloads securely, reliably and cost-efficiently on Google Cloud. At T-Digital by Deutsche Telekom I architected and operate the company's first production GCP platform — multi-region, active-active GKE that sustained FIFA World Cup 2026 peak load and passed an independent DDoS resilience test — and the AI services that run on it.
My AI work is production-adjacent and governance-focused: grounded enterprise search and RAG on Vertex AI Search with PII controls and IAM scoping; LLM guardrails and automated evaluation (toxicity, hallucination, jailbreak coverage; RAGAS); observability and SLOs for AI services; and EU AI Act / GDPR / ISO 42001 governance — gap analyses, DPIA templates, model inventories and logging controls. Two first-place finishes at Deutsche Telekom internal AI hackathons (RAG document pipeline; Jira/LLM reporting automation).
The foundation is the platform: Terraform/OpenTofu, Argo CD GitOps, least-privilege IAM and Workload Identity, Cloud Armor WAF, Cloud KMS/Secret Manager, Kyverno policies, OpenTelemetry/Prometheus/Grafana. I make AI workloads inherit the same security, reliability and cost discipline as everything else in production.
I am not positioning as an ML researcher or model-training engineer; I am the engineer who makes the AI stack safe to run. Google Cloud Ambassador (Infrastructure), Google Cloud Security Champion, 5× Google Cloud Professional, Google Cloud Technical Expert (Vertex AI, Intelligent Search, Customer Engagement Suite), Gemini Enterprise Partner Specialist. Speaker at GDG Münster on evaluating production-ready RAG systems with Vertex AI.
Email
LinkedIn
GitHubT-Digital by Deutsche Telekom · Full-time
Sep 2024 - Present · 2 yrs 2 mos
Thessaloniki, Central Macedonia, Greece · Hybrid
T-Digital by Deutsche Telekom · Full-time
Jun 2022 - Sep 2024 · 2 yrs 4 mos
Thessaloniki, Central Macedonia, Greece · Hybrid
Deutsche Telekom IT RUS · Full-time
Sep 2021 - Jun 2022 · 10 mos
Moscow City, Russia · On-site
Wildberries · Full-time
Jan 2019 - Aug 2021 · 2 yrs 8 mos
Moscow City, Russia · On-site
Mostransavto · Full-time
Jul 2017 - Jan 2019 · 1 yr 7 mos
Podolsky District, Moscow, Russia · On-site
Freelancer · Freelance
Jan 2015 - Jul 2017 · 2 yrs 7 mos
· Remote
A Chrome & Safari extension that keeps secrets out of AI prompts — it scans messages before they reach ChatGPT, Claude or Gemini using 110+ regex patterns plus Shannon-entropy analysis, then redacts or blocks API keys, tokens and PII. 100% local, zero telemetry.
Chrome Web StoreA quantitative model that gives leadership measurable visibility into key-person dependency, decision bottlenecks and human control-bypass risk — Bus Factor, Decision Concentration and Bypass Risk in one product. Published to PyPI with a citable DOI.
View on GitHubA Chrome extension that turns everyday browsing into language learning — instant translation across 39 languages, a personal vocabulary and Leitner spaced-repetition flashcards with TTS and a side-panel mode. 100% local storage, no accounts, no tracking.
Chrome Web StoreA VS Code extension that kills "works on my machine" — one-click snapshot of runtimes, Docker, ports, env vars and git state, diffed against a teammate's and turned into a Markdown fix report with ready-to-run commands. 100% local, no telemetry.
View on Marketplace
Credential ID: 8a9ddfba001e4a55bf42667a6b62da9b
Skills: Cloud Solution Architecture, Security and Compliance, Cloud Networking, +6 more
Show credentialCredential ID: 4f3e7970bbdb4a61805a8209cfe215ed
Skills: Data Engineering, Big Data & ML Pipelines, Data Governance & Security, +6 more
Show credentialSkills: Machine Learning Model Training, AutoML Workflow, Custom Model Deployment, Vertex AI SDK Usage, Prompt Engineering
Show credentialSkills: Semantic Search, Query Understanding, Information Retrieval Optimization, Relevance Scoring, Domain-Specific Search Tuning
Show credentialSkills: Conversational AI Design, Omnichannel Engagement, Real-time Agent Assistance, Generative AI Integration, Conversational Insights Analytics
Show credentialCredential ID: 9dbfab8220aa43c18da297038a95d723
Skills: CI/CD, SRE, Infra Automation, +6 more
Show credentialCredential ID: eb748cbd0c904f8389d3f755f3a32584
Skills: Cloud Networking, Hybrid Connectivity, Network Security, +6 more
Show credentialCredential ID: c658b9bd545749a6b45f54c9982f200d
Skills: Security Operations, Incident Response, Threat Detection, SIEM, +6 more
Show credentialSkills: Agent Development Kit (ADK), AI Agents, Gemini Enterprise, Memory Management
Show credentialSkills: Gemini Enterprise Deployment, AI Agents, Gemini Enterprise, Software Development
Show credential
Skills: Cloud Cost Optimization, Rightsizing, Tagging Strategy, +3 more
Show credential
Credential ID: D73996EEB4004D47
Skills: Azure Data Services, Database Fundamentals, Big Data
Show credential
Completion certificate for the NIS-2 Deep Dive web training (28 May 2026) covering real attack scenarios (identity theft, compromised backups, user account breaches), NIS-2 requirements including risk management, reporting obligations, and supply chain security, governance structures, security roles, policies, incident response processes, vulnerability management and cyber resilience with IBM solutions, network and firewall security, SOC/SIEM/endpoint protection, and identity/access/communication protection.
ACE ID IBM-0032 (v1), 14 weeks (~200 hours), minimum passing score 70. ACE: 18 SH (6 x 3) across Web Development (HTML/CSS/JavaScript), Python for Data Science, Advanced Application Security, Intro to Cloud Computing, Intro to Python, and Microservices.
Comprehensive course covering the full spectrum of working with Anthropic models through Google Cloud's Vertex AI: authentication, model selection, prompt engineering, streaming, tool use, and production integration patterns for Claude on GCP.
Show credentialComprehensive course covering the full spectrum of working with Anthropic models using the Claude API: prompt design, system prompts, multi-turn conversations, tool use, streaming, vision, embeddings, and production best practices.
Show credential7-course professional certificate demonstrating AI fluency across brainstorming, research, communication, content creation, data analysis, and coding. Includes a portfolio of 20+ AI artifacts and a vibe-coded custom AI solution. Covers prompt engineering, output evaluation, and responsible AI practices for real workplace challenges.
Show credential
Awareness eLearning Certificate of Completion covering ISO/IEC 23894:2023 Guidance on AI Risk Management: principles, context, identification, assessment, and mitigation of risks in AI systems for trustworthy and responsible AI.
Awareness eLearning Certificate of Completion covering ISO/IEC 42001:2023 Artificial Intelligence Management Systems: key concepts, structure, controls, governance, and jurisdiction-specific considerations.
Show credential
Compliance- and risk-focused training for conversational AI. Covers the EU AI Act, NIST AI RMF, and ISO/IEC 42001:2023 requirements; methods to identify, analyze, and mitigate ethical, technical, and regulatory risks; and practices for trustworthy, responsible AI implementations.
Show credentialFoundations of GDPR and privacy program governance: data mapping & ROPA, DPIA, data subject request workflows, incident response and policy design. Includes shareable certificates.
Professional, non-credit certificate covering cybersecurity fundamentals: security principles, access control, network security, incident response, business continuity and disaster recovery, and security operations.
Core concepts of OpenTelemetry for metrics, traces and logs: SDKs and auto-instrumentation, context propagation, Collector pipelines, processors and exporters — with practical integration into modern backends.
Show credential
Digital badge and certificate issued by Cloud Assess Academy for completing the Advanced GDPR course, providing advanced understanding of GDPR compliance, data protection principles, and organizational obligations.
Show credential
Intro course and exam covering ISO/IEC 27001:2022 fundamentals: ISMS scope & context, Annex SL structure, Statement of Applicability (SoA), mandatory controls, audit program basics and step-by-step implementation guidance.
Show credential
Certificate of Completion for successfully completing the ISO/IEC 27701 Data Protection and Privacy Information Management course, covering the extension of ISO/IEC 27001 for privacy management, GDPR-related practices, and implementation of a Privacy Information Management System (PIMS).
Hands-on introduction to SBOM standards (SPDX/CycloneDX), artifact signing and attestations (Sigstore/Cosign), and CI/CD automation patterns to secure the software supply chain.
Show credentialZero Trust fundamentals with SPIFFE/SPIRE; identity-based authN/Z, microsegmentation and policy design.
Show credentialIntroductory badge focused on the OWASP Top 10. Covers key web application security concepts and common vulnerabilities, risk assessment and mitigation, secure coding practices, and the confidentiality–integrity–availability (CIA) triad; includes guidance on applying appropriate security measures.
Show credentialPractical course on designing and deploying microservices and serverless applications. Covers service decomposition, API design, event-driven communication, containers and Functions-as-a-Service, API gateway and messaging basics, CI/CD and observability. Includes hands-on labs that build a small microservice with serverless triggers and deploy it to the cloud.
Foundational training in observability with Datadog: metrics, logs, traces (APM), dashboards and monitors, SLI/SLOs, alerting and integrations. Introduces tagging strategy, service maps, synthetic tests and basic Real User Monitoring (RUM). Guided exercises instrument an app and build actionable dashboards and alerts.
Professional Certificate program covering Python scripting, automation of system administration tasks, configuration management, version control with Git, troubleshooting and debugging, and using cloud APIs to automate IT infrastructure at scale.
Engineer's Degree, Automation Management Systems
Additional qualification: Translator in the Field of Professional Communication (English) — awarded alongside the Engineer's Degree
Engineer's Degree recognized by the KMK Central Office for Foreign Education (ZAB) as equivalent to a German Bachelor's degree
Time Management · Self-discipline · Communication and Collaboration
Bachelor's Degree, Law
Data Privacy and Security · Compliance and Regulations · Communication and Collaboration · Technical Analysis
8publications onGoogle Cloud - Community · Data Engineer Things · Medium · Mr. Plan ₿ Publication · Code Your Own Path
Google Cloud - CommunityMay 14, 2026
How to security-test LLM systems: red-teaming for prompt injection, detecting data leakage, and validating tool-call safety — building an AI security evaluation aligned with responsible AI.
Show publicationGoogle Cloud - Community
Architecting agentic AI with the Agent Development Kit (ADK) and skills: tool use, planning, memory and orchestration patterns for moving from ad-hoc prompts to production-grade agents.
Google Cloud - Community
A minimum security baseline for Vertex AI: IAM and Workload Identity, VPC Service Controls, CMEK, private endpoints, and guardrails to run generative AI safely on Google Cloud.
Data Engineer Things
Comparing RAG architectures for 2026 — naive, hybrid, reranking, graph and agentic RAG — with guidance on when each pattern fits accuracy, latency and cost requirements.
Medium
Three cautionary tales of ML in production: training/serving skew, data drift, and infrastructure mismatch — and the MLOps practices that keep models reliable beyond the notebook.
Mr. Plan ₿ Publication
A practical framework for evaluating LLM products: offline vs online metrics, golden datasets, LLM-as-judge, regression testing, and building an evaluation loop you can actually trust.
Data Engineer Things
End-to-end guide to building a Retrieval-Augmented Generation system: chunking, embeddings, vector search, retrieval and generation, with practical trade-offs for production RAG pipelines.
Code Your Own Path
How Mixture-of-Experts architectures scale model capacity without scaling inference cost: sparse activation, expert routing, and why MoE powers today’s most efficient large models.
4distinctionsselected by Google & the developer community
One of 41 practitioners worldwide selected by Google Cloud as Ambassador for Infrastructure — listed in the official Google Cloud Ambassadors directory. Awarded for sustained, hands-on mastery of Google Cloud infrastructure and community impact, proven in production rather than on paper.
Recognised by Google as a Security Champion — a distinction given to practitioners who consistently demonstrate deep, applied expertise in securing Google Cloud environments and raising the security bar across the community.
Won 1st place at the 3rd RIL AI Hackathon, hosted by the Research Innovation Lab and associated with T-Digital (Deutsche Telekom). Our team built a production-ready data uploader tailored for RAG-based chatbots. It supports PDF document ingestion, offers multiple adaptive chunking strategies, works autonomously with uploaded corpora, and automatically selects the best strategy using LLM-driven evaluation. We also instrumented RAGAS for response-quality measurement.
Won 1st place at the 2nd RIL AI Hackathon, hosted by Research Innovation Lab and associated with T-Digital (Deutsche Telekom). Developed an AI solution that analyzes user stories and test cases, providing insights to improve quality and efficiency in software development.
