planned
Arthur Zakirov
By day, I build production AI systems at a World Economic Forum-recognized AI scale-up, where I’ve helped 20+ engineers and leaders adopt agentic tooling.
Outside work, I use the bleeding edge to automate my own life — then turn what works into guides and tools you can use.
📚 Guides from Real Accomplishments
published
I Found a 770-Hour Automation Opportunity Hidden Behind a 41-Hour Roadmap. Senior Leaders Started Reaching Out. Here’s How You Can Spot the Same Pattern.
published
1,000+ Hours of Using Agentic Assistants in 10 Minutes
published
The Human Context Compiler: How to Turn Messy Business Intent Into Agent-Executable Work
published
I Went From My First Line of Code to Teaching 20+ Senior Engineers AI Workflows in 2 Years. These Are the 7 Adaptability Cheat Codes I Used.
published
I Cleared 100+ Tickets in 2 Days When a Senior-Led Automation Effort Had Stalled for Weeks. Here's How.
🛠️ Tools You Can Use
Practical systems I use to get real work done. You can use them yourself by opening the links below and following the setup guides.
AgentDesk
Make any machine ready for human and AI-agent work.
OpportunityOS
Turn high-volume applications, registrations, and browser searches into repeatable opportunity workflows.
SystemSmith
Engineering methods for code, agents, and architecture.
ProofStack
Turn real achievements into persuasive, public-safe career proof and positioning.
SkillPort
Package agent skills once. Install them anywhere.
✅ Results & Track Record
The outcomes, experience, and education behind the guides and tools — so you can judge whether these methods lead to results you would want.
KONUX
Junior Data Engineer - Agentic AI, Data & Cloud Platform
Nov 2025 - Present
KONUX is a Munich AI scale-up applying predictive maintenance to rail infrastructure, recognized by the World Economic Forum as one of the world’s leading technology start-ups. I started with legacy Java service and database migration work. Through proactive ownership, my role expanded into production agentic AI systems, engineering enablement, and senior-stakeholder consultation on high-leverage automation opportunities.
- Helped maintain team velocity after the team shrank from 10 to 8 people while inheriting Technical Delivery responsibilities from 3 additional engineers, by building tooling, executable SOPs, and workflow coverage.
- Helped secure ISO27001 recertification by clearing 100+ CVE tickets in 2 days using local agentic skills and API-connected workflows, after a 2-senior-engineer, multi-week automation effort had not solved the audit bottleneck.
- Built and deployed the company’s first production agentic-AI automation with LangChain, Pydantic, Databricks, Delta Lake, and MLflow, automating Jira-ticket ingestion, diagnosis, PR creation, Slack notifications, and Jira status updates.
- Shifted the automation roadmap from 41 to 770 hours of saveable work — ~19× more leverage — by building a Databricks pipeline from IT Service Management data and visualizing operational bottlenecks for leadership.
- Became a direct technical contact for senior stakeholders, including the Head of Engineering and Chief of Staff, on operational KPIs, automation priorities, practical AI tooling, and productivity workflows.
- Moved ~2 months of planned engineering work forward in ~2 weeks across workflow automation, operational tooling, and on-call onboarding using local agentic workflows connected to AWS Cognito, DynamoDB, EventBridge, Databricks, Bitbucket, Jira, and Slack.
- Equipped ~20 engineers with local agent harnesses, API tool-calling across company systems, and procedural/business-knowledge skills, using a Dockerized FastMCP server proxy architecture.
- Ran 4 AI-tooling workshops and targeted 1:1 onboardings for engineers, senior leaders, leads, and a principal engineer, driving practical AI adoption across engineering and receiving CEO recognition.
- Exceeded expected month-6 velocity in month 1 on legacy Java service development in a hexagonal architecture, despite no prior Java experience.
KONUX
Data Science Intern - Fullstack ML Training Data Apps
Apr 2025 - Sep 2025
Earlier, during my internship, I worked in a cross-functional data science team, building internal fullstack apps and evaluation workflows around ML training data.
- Diagnosed misinformed model deployments and built a near-production-grade Python evaluation framework that was adopted team-wide, saving ~2 days/week of potentially wasted annotation effort.
- Built a Python Dash app automating ETL with Jupyter, CLI, AWS S3, and Databricks, replacing ~2h manual daily setup with 10-min UI clicking for 3 Senior Data Scientists.
- Migrated an on-prem app to cloud with S3, Athena, ECR, Kubernetes, Docker, Poetry, Bash, and Python, enabling 3+ team members to scale AI model training.
Tacto
AI/ML Engineering Intern - Technical Solution Engineering
Nov 2024 - Apr 2025
At a €55M-funded BME Procurement Excellence Award-winning startup, I worked in a dynamic, high-pressure, multi-stakeholder environment demanding resilience, multitasking, and rapid prioritization.
- Migrated customer data pipelines from Prefect to Dagster, reducing tech debt and enabling product teams to deliver new ERP analytics features.
- Diagnosed and resolved data quality issues in staging and production using SQL and Python to trace root causes.
- Cleaned and structured raw ERP datasets from new customers using Python scripts, prompt engineering, and Google Sheets to support smooth technical onboarding.
- Built an internal Retool app to streamline Azure data setup, automating multi-step workflows into a 5-click UI.
Arthur Zakirov
Projects - Fullstack AI Data Apps
Apr 2024 - Nov 2024
Between roles, I used self-directed projects to build practical full-stack data apps around problems I personally understood and could validate end to end.
- Independently built full-stack MVP apps using Python, data pipelines, dashboards, and automation.
- Created a full-stack nutrition planner using web scraping, LLMs, and mathematical optimization to personalize meals by 20+ metrics such as cost, prep time, satiety, fat loss, and micronutrients.
- Integrated OpenAI, LangChain, fuzzy matching, and RAG-style retrieval to unify disparate nutrition datasets.
- Designed interactive dashboards with Plotly to visualize trade-offs, nutrient intake, and goal alignment.
- Engineered an automated financial tracking tool aggregating transactions from 6+ platforms, enabling cash flow tracking, P&L analysis, and tax preparation.
EngineeringAthlete
Founder - Generative AI, Content Creation, Consulting, Courses
Oct 2022 - Sep 2023
EngineeringAthlete was a one-person online business where I combined content, coaching, generative-AI services, digital products, automation, sales, and operations for an international audience.
- Delivered coaching and digital self-improvement tools to 120+ international customers, generating up to €5,800/month in revenue.
- Grew an English-language YouTube channel to 60,000+ subscribers and 1M+ views by scripting, filming, editing, and publishing videos solo.
- Created and sold end-to-end service packages combining fitness programs, AI-generated profile photo enhancement, and educational content for improving personal appearance.
- Set up and automated landing pages, payment, onboarding, invoicing, and access control for a seamless customer journey.
- Handled all customer communication, sales, copywriting, and operations — including accounting and legal — in English.
Bosch Center for Artificial Intelligence
AI Enablement Intern - ML Research Implementation
Jan 2022 - Mar 2022
Bosch is ranked by Fortune as the No. 1 vehicle-parts supplier in the world. At the Bosch Center for Artificial Intelligence, I worked on AI/ML enablement and research implementation.
- Developed AI/ML training materials using Python, PyTorch, and NumPy by implementing and evaluating algorithms such as semi-supervised and active learning on limited datasets.
Technical University of Munich
Master of Science - MS, Robotics, Cognition, Intelligence
2022 - 2024
TUM is ranked by QS as one of the best engineering faculties in the world. QS ranking article
- Grade: discontinued in favor of practical experience.
- Studied for one semester before shifting focus to self-directed projects more closely aligned with real-world market needs.
- Machine Learning from Scratch reading/project list: GitHub list.
- Relevant lectures included Introduction to Deep Learning, Machine Learning, Machine Learning for Graphs and Networks, Reinforcement Learning, and Fundamentals of Artificial Intelligence.
Technical University of Munich
Bachelor of Science - BS, Mechanical Engineering
Oct 2018 - Sep 2022
TUM is ranked by QS as one of the best engineering faculties in the world. My overall grade was achieved by only 12 students out of 683 degrees awarded in total, according to the grading table — approximately 1.8%. QS ranking article
- Chose a research topic originally intended for a Master’s thesis and self-studied 6 graduate-level AI/ML courses to support it. Thesis grade: 1.3.
- Bachelor’s Thesis: Sensitivity Analysis of Generalizability of Pattern-based Prediction Methods for Autonomous Driving.
- Thesis repository: Bachelor thesis generalization analysis repository.
- Improved generalization of deep learning-based motion prediction algorithms in autonomous driving by developing a custom data augmentation process.
- Performed sensitivity analysis on model performance to assess prediction robustness across new, unseen environments.
💬 What Others Say
“His performance has received high recognition from us […] He can process information quickly and is easily able to take complex cross-divisional interrelationships into account. […] Within his area of responsibility he always dealt with problems independently, offered constructive alternatives and understood how to implement these independently when required. Across all subject areas, he achieved very good results in both qualitative and quantitative terms. […] Arthur Zakirov impressed us above all with his great social competence.”
“Herr Zakirov zeigte stets eine ausgezeichnete Leistungsmotivation, Eigeninitiative sowie Einsatzbereitschaft […] Er arbeitete sich […] innerhalb kürzester Zeit in neue komplexe Aufgaben ein. […] Die Arbeitsergebnisse waren […] immer von herausragend guter Qualität. […] Herr Zakirov hat mit seinen Leistungen stets und in jeder Hinsicht unsere höchste Anerkennung gefunden.”
“An der Stelle auch ein großes Danke an dich! Ich bin beeindruckt von deiner Entwicklung und deinem Einsatz. Der Austausch hilft auch mir persönlich, Dinge nochmal zu hinterfragen und mich weiterzuentwickeln. Ich denke ich spreche für uns alle, wenn ich sage, dass wir uns sehr freuen dich an Board zu haben!”
“Your enthusiasm and drive to learn new skills is outstanding.”
“Your enthusiasm and work spirit has been quite inspiring.”
“You are one of the most inquisitive people I've met. Don’t lose that.”
🏋️ Beyond Automation: Strength & Wisdom
The same pattern shows up outside my work: pick a hard domain, study it seriously, practice consistently, and turn the result into judgment.
I lift heavy weights
At my physical peak, I reached 1.80 m, 84 kg, roughly 9% body fat, and a 170 kg × 10 deadlift while finishing mechanical engineering at TUM with distinction. My estimated fat-free mass index was around 23.6–23.8, a level that Kudsk et al. report near the 99th percentile in their population table. That combination led me to start EngineeringAthlete: an attempt to merge rigorous engineering thinking with high-performance fitness. Today, I train sustainably, but I’ve moved the deeper intensity into one main craft: agentic automation.
Kudsk et al. FFMI table
I chase wisdom
There was a time when I learned primarily by sitting down with books. Today, most of my idea intake comes from audiobooks and podcasts during the parts of life where I am away from the computer. When I do have focused time, I usually choose the computer: building, testing, writing, and turning knowledge into something tangible with a tighter feedback loop.
🤝 Ways to Get in Touch
Reach out if you are trying to turn messy workflows into practical AI or data systems, clarify an automation opportunity, or apply one of the methods from the guides.