Platform engineer at A1 Bulgaria, building the internal developer platform on
Kubernetes, Crossplane, and Kyverno. Trained as a software engineer first: Go controllers on
the platform side, full C#/.NET applications on the product side.
Came up through enterprise identity and Azure support, which is why security
and auditability get designed in early, not bolted on later. Now extending the same platform
patterns toward AI infrastructure: GPU scheduling and model serving on Kubernetes.
Open to platform engineering, software engineering, and AI infrastructure roles
at ambitious, international engineering teams.
STUDY
Deliberate, independent study, organized as topic overview notes that fan out
into dedicated roadmap notes, not just what the job requires. Active right now:
HashiCorp ecosystem:Vault internals (secrets engines, seal and unseal, dynamic credentials, auth methods), plus Consul, Nomad, Packer, and Sentinel policy as code.
DevSecOps, security by design:a self directed track through appsec, container scanning, Kubernetes hardening, and compliance as code (OPA, tfsec, CIS, NIST), building toward ITIL, Azure Administrator, and DevSecOps Professional certifications.
Cloud native platform engineering:Kubernetes and Docker internals beyond Crossplane's abstraction layer, plus CI/CD across Jenkins, GitHub Actions, and GitLab CI, and observability with Prometheus, Grafana, and the ELK stack.
Go and Python fundamentals:goroutines, interfaces, and structs in Go; OOP, file I/O, and regex in Python; plus algorithm practice, on the way to AI infrastructure work.
System design and OOP modeling:entity and class design with defined attributes and cardinality, worked out for Aetherion rather than just sketched as a product idea.
Extended a Crossplane provider (Go) to add storage encryption, replication across regions, and automatic drift correction.
Designed configurable etcd leader placement across a Kubernetes fleet spanning multiple zones, coordinated across CRD, controller, and Helm chart.
Built an automated QA pipeline testing every weekly Crossplane release across customer environments before deployment, using Kyverno Chainsaw, GitHub Actions, and self hosted ARC runners.
Built a VM snapshot scheduler on Exoscale's API, automated through Crossplane on a fixed schedule.
Designed OIDC login to Vault via IdP group claims, with least privilege policies per application, managed declaratively through Crossplane.
Found ArgoCD was authenticating to clusters with full admin access, bypassing Kyverno entirely; fixed it with a scoped ServiceAccount and token per cluster instead.
2024/25
SEGA :: Junior IT Technician
Internal IT systems, hardware/software troubleshooting.
2022/23
TeKnowledge :: Technical Support Engineer, Stage 2
Outsourced Microsoft Azure AD/Identity support using PowerShell, Kusto, and Fiddler on hybrid/cloud issues.
2023
Fontys University of Applied Sciences :: ICT & Software Engineering
Propedeuse, 9/9, Eindhoven. Bachelor's completion under consideration.
A movie streaming platform built end to end in C#/.NET: a Windows Forms
desktop client and a Razor Pages web app sharing one backend, backed by a SQL Server
catalog of 3,000+ titles with recommendations, watchlists, and viewing history.
C#.NETRazor PagesWindows FormsSQL Server
Why this shape
One shared backend and database behind two front ends, desktop and web, instead of duplicating logic per client.
Full catalog, trailers only: the movie data is real, playback isn't, since the project never had streaming rights.
Led a team building a role based logistics management system in C#/.NET:
stock tracking, task allocation, and admin reporting across four separate roles, from
warehouse security to department management.
C#.NETSQL ServerTeam Leadership
Why this shape
Led the team on scope and integration, not just one module, coordinating four role based views into one system.
Access scoped by role (security, depot, sales, management) instead of one generic admin panel, matching how the business actually splits responsibility.
An AI learning platform inspired by Star Wars holocrons: name any skill and
it builds a path of short trials, each a streamed lesson plus a quiz, with XP, ranks and a
tutor chat. Accounts and progress persist per learner.
Next.jsReact Three FiberVercel AI SDKGroqSupabasePostgres RLSVercel Cron
The model layer reads Groq's live model list and falls back across models on rate limits or retirements, so the AI keeps working without manual changes.
Authorization lives in Postgres row-level security, not just the app, and the AI routes reject anonymous calls so only signed-in learners spend the quota.
A daily Vercel Cron job keeps the free database from pausing and health-checks the AI.
A homelab run entirely through GitOps: automated k3s provisioning on
Vagrant/VMware, workloads reconciled by ArgoCD, access scoped through RBAC, monitoring,
and remote access locked down over Tailscale.
k3sVagrant/VMwareArgoCDRBACTailscaleMonitoring
Why this shape
k3s over a full RKE/managed cluster, to keep the homelab's footprint small enough to rebuild from scratch.
Tailscale for remote access instead of exposing anything publicly or running a VPN appliance.
Reproducible DevOps/security VM environments across Windows and macOS
hosts, using Vagrant, VMware, and Ansible to provision Ubuntu and Kali boxes on demand.
VagrantVMwareAnsibleUbuntuKali
Why this shape
One Vagrantfile and playbook set targets both Windows and macOS hosts, instead of maintaining two setups.
Kali alongside Ubuntu turns the same lab into a disposable security testing environment on demand.
SEE ALSO
Hobbies:produced original samples and instrumentals in FL Studio and ran a small beat store years ago; more recently, 3D art in Blender. Also nature and cycling.
Tools:iTerm, VMware Fusion, Proxmox (homelab, in progress), and OrbStack for virtualization; Tailscale for homelab and VPN; Figma and draw.io for design and diagrams; Obsidian for a Zettelkasten built up over a few years; Anki for language learning.
GEAR
The actual setup and tools behind all this, for anyone who cares about that sort of thing.
