Tacconi Consulting

Tacconi Consulting Ltd · London

Platform engineering for teams that run their own infrastructure.

Kubernetes on bare metal and in AWS, GitOps, and virtualisation that does not depend on a VMware renewal. The practice is led by Riccardo Tacconi, a platform engineer with 26 years in infrastructure and the founder of the kcore hypervisor.

Start a conversation Selected work

Services

Three things we are hired for

  • Platform engineering

    Design, build and operate Kubernetes platforms: bare metal with Talos Linux, AWS EKS, or Proxmox. GitOps with Argo CD, Terraform for everything that has an API, CI/CD for infrastructure, and the day-2 work that decides whether a platform lasts: upgrades, storage, networking, runbooks.

  • Leaving VMware or Proxmox

    An assessment of the estate first, then a migration onto kcore with a plan you can hold us to. We inventory the source hypervisor, convert disks with kcore-migrate, and bring Windows and Linux guests across. If kcore is not the right target we say so before any work starts.

  • Modernisation with AI assistance

    Legacy estates analysed, specified and re-platformed with agent-assisted workflows, kept under tests, review and CI gates. Generated code is reviewed by an engineer before it merges. The result is a codebase your team understands, not a black box.

Engagement detail and how we scope work

Products

kcore

We build and maintain kcore, an open-core virtualisation platform for edge sites and datacenters. It grew out of problems we kept meeting in client work: hypervisors with an API bolted on afterwards, host updates that cause outages, and licences priced per core. The product line has three parts.

kcore hypervisor

KVM virtualisation on immutable NixOS hosts. The gRPC API is the primary interface; kctl, the Terraform provider and the MCP server are clients of it. Host updates are one transaction and roll back to the previous generation if they fail.

Windows and Linux guests, GPU passthrough, Ceph shared storage with live migration, mTLS throughout. Priced per CPU socket; the evaluation ISO is free.

kcore Vault

Off-cluster backup for kcore volumes. Snapshot-based with changed-block tracking, so only the extents that changed since the last run leave the cluster. Data moves over mTLS under a separate trust root, so a compromised cluster cannot read its own backups.

Shipped on the private channel for paying customers. In active development.

kcore Provisioning

Bare-metal provisioning for kcore fleets. Discovers and controls servers over Redfish and IPMI, netboots hosts with UEFI HTTP boot, verifies every image by SHA-256, and issues host certificates from its own PKI.

Written in Rust, with mTLS, audit logging and a read-only dashboard.

Alongside these: kcore-migrate for VMware and Proxmox inventory and disk conversion, a Terraform provider, an MCP server for AI agents, and a Crossplane provider. About the kcore product line · kcorehypervisor.com

Selected work

Recent engagements

  • Automata · Staff Platform Engineer · 2025–2026

    Edge-to-cloud platform for lab automation

    Centralised control plane on AWS EKS managing a global fleet of edge Kubernetes clusters. Temporal workflows in Python for provisioning, lifecycle and remediation; Talos Linux at the edge; Terraform and Argo CD throughout. Owned the security model and operational standards end to end.

  • Crystal Intelligence · Senior Platform Engineer, interim Infrastructure Lead · 2024–2025

    Kubernetes on bare metal and Proxmox

    A production Kubernetes platform in an environment without cloud primitives: Talos Linux for immutable hosts, LINSTOR for replicated block storage with automatic failover, and a restructured AWS estate with clearer security and billing boundaries.

  • Builder.ai · Lead Platform Engineer · 2023–2024

    Fleet operations at roughly 3,000 clusters

    Operated and improved the internal platform that provisioned and managed Kubernetes clusters across AWS and Azure for AI-driven applications, including the GitLab CI/CD pipelines that deployed to them.

  • BBC · Senior Systems Engineer · 2020–2021

    Monitoring platform replacement

    Replaced the Zenoss monitoring estate for broadcast and IT infrastructure across AWS, VMware and on-prem Linux with an open-source stack, automated with Ansible and Terraform. The consolidation saved the BBC several million pounds.

Full record of work, 1999 to date

How we work

Short discovery, written decisions, clean handover

Engagements start with a short discovery: what exists, what is breaking, what the constraints are. From that we propose a target design and an incremental path to it, in writing, so you can disagree with it before anything is built.

Everything lives in Git and runs through CI from the first week. Architecture notes, runbooks and decision records are deliverables, not afterthoughts. At handover your team operates and extends the platform without us; if it cannot, the engagement is not finished.

Where AI tooling is involved, the same rules apply: tests first, a human reviews every change, and nothing merges without passing the gates.

Contact

Start a conversation

Tell us what you are trying to do and what is in the way. For a VMware or Proxmox exit, include socket count and the rough mix of Windows and Linux guests; we reply with whether kcore is a sensible target and what the project would look like.