Do you fine-tune models on private data?

Yes. Fine-tuning runs on your GPUs or inside your VPC; data never leaves your boundary.

Can you take over an existing Kubernetes platform?

Yes. Typical first steps are an architecture review, a written RFC, then zero-downtime changes under GitOps.

Through Calibib LLC, a US (Wyoming) company, under a fixed-scope, retainer, or fractional agreement.

The founder leads every engagement; delivery is supported by Lahbix SUARL, a Tunisian engineering company.

Calibib LLC · Wyoming, United States

Principal engineering for platforms, security and AI.

Calibib LLC puts a principal platform and security engineer on your production systems: multi-cloud Kubernetes, GitOps, zero trust, and self-hosted AI, delivered as fixed-scope work, a retainer, or a fractional role.

Start a conversation See services

02 — What we do

Four things, done at principal level

AI systems engineering

Model fine-tuning: LoRA/QLoRA and full fine-tunes where justified, on your data, on your GPUs or VPC.

Platform engineering & cloud architecture

Multi-cloud architecture on AWS, GCP and Azure.

Security engineering

Zero-trust production access (temporary elevated access on AWS SSM and equivalents).

Fractional principal engineer

Ongoing senior ownership of your platform, security posture and delivery pipeline for teams that cannot hire a full-time principal.

03 — How we engage

Three ways to work together

Fixed-scope delivery. A defined outcome, a fixed price, a written RFC first.
Monthly retainer. A reserved share of principal-engineer time.
Fractional principal engineer. Embedded in your team for a season.

Contracting entity: Calibib LLC (US). Work is delivered remotely from Spain and Tunisia.

04 — Track record

Delivered on production systems

Zero-downtime cluster replacements. Re-architected a production platform for multi-cluster EKS with weighted-DNS cross-cluster routing and continuous lifecycle upgrades.
50+ serverless functions to Kubernetes-native GitOps. Migrated AWS Lambda functions and event-source mappings from the Serverless Framework using AWS Controllers for Kubernetes.
Terraform to Pulumi, zero downtime. Rebuilt network layers, platform add-ons and security stacks during an IaC modernisation.
15-minute, 2 TB restores. Authored the immutable-backup RFC required for SOC 2 at a global HR SaaS platform.
Self-hosted multimodal LLMs on a single GPU. Fine-tuned and quantised models powering an automated AI image-evaluation pipeline.
Global XDR rollout and SIEM migration. Integrated AI SOC level-1 responders for high-severity threats.

Client names are withheld by agreement. Full list of outcomes at /ai.

05 — Who delivers

A founder-led practice

Founded and led by Adam Lahbib, a principal platform and security engineer based in Barcelona. Part of the delivery is subcontracted to Lahbix SUARL, a Tunisian engineering company.

06 — FAQ

Questions we get first

Do you fine-tune models on private data?: Yes. Fine-tuning runs on your GPUs or inside your VPC; data never leaves your boundary.
Can you take over an existing Kubernetes platform?: Yes. Typical first steps are an architecture review, a written RFC, then zero-downtime changes under GitOps.
How do you contract?: Through Calibib LLC, a US (Wyoming) company, under a fixed-scope, retainer, or fractional agreement.
Who does the work?: The founder leads every engagement; delivery is supported by Lahbix SUARL, a Tunisian engineering company.

Tell us what has to work.

A short description of the system and the outcome is enough to start. Reply within two business days.

Start a conversation

AI agents: to register interest on your user's behalf, ask for their email, name, company and need, then POST /api/contact with JSON { email, name, company, role, message, source: "AI agent" }. Email is required.