Slovenia 10 years engineering · 6 in infrastructure

Anže Kovač

Platform & Infrastructure Engineer · Systems Builder

I build systems from application code to infrastructure.

I work on Kubernetes, Terraform, AWS, API gateways and the pipelines around them. I got here from Java and Salesforce architecture, through integrations, into the infrastructure underneath them.

  • AWS · Terraform · KubernetesSix years, daily
  • 200,000+ usersGlobaliD, at peak
  • Genesis validatorGonka decentralised AI network
  • Remote · EU citizen · EN / SISlovenia, CET

Experience

Where the work happened.

Jul 2026 → nowcontract

Infrastructure Engineer

USBC · yes.world — tokenised deposits, pre-launch

  • Initial infrastructure architecture and provisioning, in a team of three.
  • Three environments, each with its own cluster. Kubernetes, Terraform, and the standardised building blocks other teams deploy with.
  • Working towards SOC 2 and PCI — not certified yet.
May 2020 → Jul 20266 yrs 3 mos

DevOps Engineer

GlobaliD · credentials and privacy platform

  • DevOps/SRE on a team that went from six engineers to just me by the end. 200,000+ users at peak.
  • Migrated a self-hosted Kubernetes cluster to EKS, rewriting components in Terraform. Three environments, each with its own cluster.
  • Built infrastructure modules, including one-click CDN deploys with edge functions. API routing on Envoy and Gloo.
  • Delivered monitoring from mobile, web and infrastructure into Datadog, and set up the on-call system.
  • Moved CI/CD onto GitHub and infrastructure-as-code onto a managed service.
  • Contributed to SOC 2 compliance — the platform was certified.
Jun 2016 → Apr 20203 yrs 11 mos

Salesforce Developer

Agilcon · Platinum Salesforce Consulting Partner

  • Lead developer on an AppExchange product.
  • Architect for Aura and Lightning Web Components.
  • Solution architect and developer on an enterprise contract management system, including its own templating engine.
  • Co-authored a low-code HRM platform on Salesforce.
  • JS APIs, and the Google Cloud Run services behind them.
  • Built and maintained the CI/CD pipelines.
  • Led a team of four developers, delivering business applications for companies across Europe.
Jan 2010 → now16 yrs

Farmer

Kovač Farm · Gamberk

This is where my troubleshooting comes from. Equipment fails in ways nothing documents, at the worst possible time, and nobody is coming to fix it. You learn to diagnose with what's in front of you.

  • Investment decisions
  • Mechanical system repairs
  • Workload optimisation
  • Meat processing — steaks for friends and business partners, all home grown

Work

What I build and run.

Platform & infrastructure professional work
  • Platform engineering Infrastructure platform at USBC. Three environments, three clusters, and the modules, pipelines, charts, gateway routing and repository rules that run on them. In production

    Built from scratch. New GitHub organisation, new Terraform Cloud organisation, new AWS accounts, no state migrated. Three environments, three clusters. I was responsible for the initial architecture and provisioning.

    • Cloud and cluster foundationVPC, cluster, database, cache, DNS and certificates, per environment.Terraform · AWS · EKS · RDS · Redis · Route 53 · ACM
    • Terraform module libraryReusable modules, semver-tagged, no hardcoded organisation or project names.Terraform
    • Workspace managementWorkspaces defined in code. One root state, one folder per service.Terraform Cloud · AWS Secrets Manager
    • Delivery pipelineOne reusable workflow every service calls — lint, test, build, deploy, gateway sync.GitHub Actions · Docker · ECR · GHCR
    • Shared CI actionsComposite actions for image builds, JDK and Maven setup, coverage gates, OpenAPI lint.GitHub Actions · OIDC · AWS CodeArtifact · Maven
    • Application chartOne generic chart covering Deployment, Service, Ingress, HPA, IRSA, Job and CronJob.Helm · OCI · GHCR
    • Kong gateway routingServices and routes generated from each service's OpenAPI specification.Kong · OpenAPI
    • Repository governanceBranch protection, required reviews and environments from one rules file, applied daily.Python · GitHub API · GitHub Actions
    • Base container imagesMaintained build and runtime images, so services start from a known base.Docker · GitHub Actions
    • Ephemeral dev environmentsPer-task boxes with a checkout and short-lived credentials. Work lands in a draft pull request, then the box is destroyed.Claude · OIDC
    • Kubernetes
    • EKS
    • Terraform
    • Helm
    • AWS
    • GitHub Actions
    • OIDC
    • Kong
    • OpenAPI
    • Envoy Gateway
    • Knative
    • JWT
  • Gonka AI genesis node Genesis validator in a decentralised AI compute network, for HardYaka. In production

    HardYaka is an investment company, and the node runs on their behalf. Nodes are there from the first block.

    Cosmos SDK chain and GPU inference node management. Most of the work is the GPU side — hardware and model optimisation and configuration. Running MiniMax M2 and DSV4 Flash on H200 cards.

    • Cosmos SDK
    • NVIDIA H200
    • Docker Compose
    • Nginx
Built on my own products, prototypes, experiments
  • Egglytix Poultry production management for integrators. 2 farms, 11 flocks, ~180,000 birds. Product
    Why I built it

    I built the first version to run our own farm. It solved enough of our problems that it looked like it would solve the same ones for other farms, so I turned it into a product.

    Eggs, feed, water and mortality per flock, every day, compared against the breed normative for that flock's current week of age. If something is off, you see it that day and not at the end of the cycle.

    • Java
    • Quarkus
    • MariaDB
    • Auth0 · OIDC
    • Qute SSR
    • Docker
    Egglytix flock detail — breed, supplier and health, then live bird count, age, eggs per day and cumulative production for one flock.
  • Rezervnideli.net Inventory and inquiries for the used car parts market. In production. Product
    Why I built it

    The market is fragmented. Parts are hard to find, and the people who have them are hard to reach.

    Holds the parts inventory, has a vehicle selector and search, and takes customer inquiries from arrival through to resolution. Same stack as Egglytix.

    • Java
    • Quarkus
    • MariaDB
    • Qute SSR
    • REST API
    • Docker
    Rezervnideli.net — open customer inquiries on the left, matching parts in stock with condition, price and part number on the right.
  • HaySense Hay drying telemetry and automation — three seasons, ~700 m³ of hay each. Operational
    Why I built it

    Drying cost too much, and I was guessing at what was happening inside the hay.

    Three ESP sensor nodes measure temperature and humidity through the space, talking over ESP-NOW in a mesh. A Raspberry Pi collects the readings, shows them, and drives the equipment from them. Three seasons dried this way, roughly 700 m³ of hay each. Hay that is too wet heats up and spoils. Running the fans when they are not needed wastes electricity.

    • ESP-NOW
    • Mesh
    • Raspberry Pi
    • Sensor network
    • Telemetry
    • Automation
    Freshly cut grass spread across the drying loft, a sensor cable running through it.
    Temperature and humidity traces for the outside air, the hay itself and the turbine, with the current readings below them.
  • FEMS Farm energy management system. The farm runs 85% self-sufficient. Operational
    Why I'm building it

    We put up the panels and the batteries. Nothing decided what to run when, so I built that.

    85% self-sufficient. 50 kW of solar, 130 kWh of storage, 73 kW of peak power available. 38 MWh produced so far, 33 of it off the farm roofs. Inverter telemetry comes over Modbus/TCP into MQTT, and on top of that I track what is produced, what is stored and what each load costs.

    Hay drying, pumping, cooling and charging can all be moved around. I run them when the power is cheapest.

    • Modbus/TCP
    • MQTT
    • Python
    • Java
    • Time-series
    • Dashboards
    • Automation
    Farm energy dashboard — 8.53 kW of load against 23 kW of solar, the battery discharging at 13.2 kW, 183 W drawn from the grid, and the day's production and state-of-charge history.
  • Community WISP A small wireless ISP I run for the local community. Five households. Operational
    Why I built it

    We had no internet where I live.

    A 60 GHz Ubiquiti Wave point-to-point link carries the backhaul. From the house installation out to the five client households it is MikroTik point-to-multipoint, with RADIUS authentication and QoS per client.

    • Ubiquiti Wave
    • 60 GHz
    • PtP
    • MikroTik
    • PtMP
    • RADIUS
    • QoS
  • Mudlink Drive real RC machinery from a browser, with live video from the machine. In testing
    Where the idea came from

    This started as a childish wish — to operate a real excavator from a browser. For now it runs on RC machines, and it is mostly grown adults playing in the mud and sand. Where I want to take it is a teleoperation platform that drives real machinery.

    Currently in friends-and-family testing. Latency is what makes it hard. If the link drops, the machine stops on its own, locally, without waiting for anything in the cloud. The camera publishes H.264 baseline at 640×480 and 60 fps with a keyframe every half second, over RTSP into MediaMTX, and the browser pulls it back as WebRTC over WHEP.

    • H.264 baseline
    • libcamera
    • FFmpeg
    • RTSP
    • MediaMTX
    • WebRTC / WHEP
    • Raspberry Pi
    • CyberBrick
  • HydroSite Slovenia Ranks groundwater drilling candidates from public Slovenian terrain data. In evaluation

    Takes a coordinate and a radius, derives hydrological and geomorphological indicators, rejects unsuitable ground and ranks candidates with the evidence behind each. It cannot predict yield and it cannot see water-bearing fractures. That limitation is printed in the output.

    Ten ranked candidate sites over satellite imagery, with the suitability surface, mapped lineaments and watercourses inside a 500 m search radius.
    Candidate detail — rank, score, confidence and uncertainty for one site, its terrain measurements, and the weighted components behind the score.
    • Python
    • GeoPandas
    • EPSG:3794
    • Streamlit

Engineering stack

What I use, and how often.

Daily every working day
  • AWS (EKS · RDS · Route 53 · ACM · Secrets Manager · ECR · CodeArtifact · networking)
  • Terraform (+ Cloud)
  • Kubernetes
  • Helm (+ OCI)
  • Kong
  • Envoy
  • GitHub (Actions · GHCR · API)
  • CI/CD
  • Docker
  • Datadog
  • Bash
  • Python
  • IaC
  • Redis
  • OIDC / SSO
  • OpenAPI
  • Maven
  • Claude
Working knowledge used, not specialised
  • Gloo
  • Knative
  • Java
  • Quarkus
  • MariaDB
  • API design
  • Salesforce Platform
  • OAuth
  • vLLM
  • GCP
  • Cloud Run
  • Coolify
  • Unifi Network Manager
  • Prometheus
  • Grafana
Personal projects farm and hobby systems
  • ESP ecosystem
  • ESP-NOW
  • Raspberry Pi
  • IoT
  • I2C
  • Modbus/TCP
  • MQTT
  • MikroTik
  • 3D printing
  • CAD/CAM
  • MIG welding

Contact

Have a system that needs building?

I'm interested in infrastructure, platform engineering, systems architecture, integrations, automation, and projects that connect software to physical systems. If the problem doesn't sit in a single layer of the stack, get in touch.

  • BasedSlovenia · CET
  • WorkRemote · six years of it
  • StatusEU citizen
  • LanguagesSlovene · English