IT CONSULTANT · PLATFORM ENGINEERING + APPLIED AI
Tomasz Jedynak
I design and ship systems at the intersection of platform engineering and applied AI, before handing them off.
What I do
I'm at my best in unexplored territory: a brand-new problem or an old system nobody ever got around to documenting. That's where I shape the first prototypes from scratch: build an understanding of the process, identify the problem, design the solution, build it with my own hands, and hand it off, staying on in an advisory capacity when that's useful. Routine bores me fast, so once the ground is mapped, I move on to the next unknown. On the toughest legacy incidents, I drive the diagnosis and fix from start to finish, drawing on the team's knowledge wherever I need it. Whatever repeats, I automate and turn into a tool instead of doing it by hand again.
10+ years
enterprise cloud, Big Data & platform engineering
0→1
I design it and ship it myself, not just plan it
Pioneer
uncharted territory, first prototypes
Solo / small teams
high autonomy, async mode
Stack & skills
AI / Agentic / MCP
Cloud & IaC
Kubernetes & GitOps
Data & Orchestration
Python · SQL · Bash
Flagship projects
A dependency graph AI agents can query
Teams had no visibility into how a change in one repo or package rippled through dozens of connected systems: version drift and production incidents surfaced only after the fact. I designed and built a cross-repo dependency-graph toolkit (built on Crossplane/GitOps), exposed to AI agents through my own MCP server, which lets them query impact analysis, version drift, and package propagation. It caught real production incidents before they shipped. This was one piece of a larger engagement: as part of the team building this Crossplane-based self-service platform, I owned Airflow end-to-end and brought Kafka and PostgreSQL to production-grade.
From firefighting to a queryable AI workspace
I inherited, along with one colleague, an undocumented, mission-critical enterprise integration platform. Nobody else on the team touched it, so every incident had to be solved from scratch, with no map of the system. I built up knowledge incident by incident until I became the internal expert on this platform, and turned that knowledge into an AI-driven workspace built on MCP. Ad-hoc troubleshooting turned into queryable, evidence-based operations.
Middleware and migrations that connect what didn't want to connect
I designed and built a middleware platform (Airflow, BigQuery, Cloud Composer, GCP Cloud Run) integrating local Salesforce instances with the central Master Tenant, and eventually the ERP (Workday): one platform for integration and reporting instead of another round of manually patched connections. A separate, parallel project: migrating from local CRMs to Salesforce, where I was responsible for data integration, data migration, and technical advisory.
Big Data deployed across 14 countries at once
Enterprise clients needed batch and near-real-time analytics at scale. Some environments were on-prem, high-security, air-gapped, others on Azure, spread across 14 countries. I deployed Hortonworks (HDP) and Cloudera (CDH) stacks, built ETL, real-time, and ML pipelines (Hive, Kafka, NiFi, Airflow, YARN), designed HA/load-balancing/DR architectures, and automated provisioning with Python and Ansible. I also led R&D initiatives and supported pre-sales (architecture, PoC, cost estimation).
I've led teams of up to 9 people, cross-functional, technical, always staying hands-on in solution design, architecture, and development. Player-coach, not a manager on the sidelines.
Home lab / proof-of-craft
A hobby turned into a career
My whole career in IT started as a passion and it's stayed one. After hours, I run a three-node HA Proxmox cluster with GitOps via ArgoCD on K3s, and a GPU-accelerated local LLM powers my local processes and agents. One of them is OpenClaw, my personal agent that helps me with everyday tasks, drawing on its own knowledge base, backed by n8n automations. I automate my home with Home Assistant, the same instinct that led to zgranydom.pl, my second business, hosted in this same home lab. It's proof that I can build something from zero, on my own, with no brief and no one pushing me, and take it all the way to a working, production state. Not every project ends in a spectacular win, but it's working proof, not just a line on a CV.
HA Proxmox cluster
3 nodes, GitOps
K3s + ArgoCD
self-hosted GitOps at production quality
Local LLM (GPU)
powers local processes and agents
OpenClaw
personal agent, helps with everyday tasks using its own knowledge base
n8n
workflow automation
Home Assistant / home automation
a hobby turned into a business (zgranydom.pl)
How I can help
You'll probably recognize yourself if:
Contact
Let's talk
Got unexplored territory nobody's set foot on yet? Get in touch, I'd love to hear about it.
Looking for custom-tailored building automation instead? Check out zgranydom.pl.