ArnTechCloud
Talk about your project

AI guidance · Agent operations · Cloud · Security

Your AI agents deserve real infrastructure.

I guide companies from strategy to daily operations: designing, securing, deploying and running their AI agents in the cloud, on-premises or locally, drawing on over 25 years of experience in infrastructure, messaging and security.

Talk about your project Explore the guidance
  • 25+ yearsin IT infrastructure
  • On-prem and cloudAzure, AWS, Microsoft 365
  • Security and messagingidentity, access, email
  • Cloud and local AIagents, RAG, local models

Why it matters

An agent doesn't just answer. It acts.

Once an AI can read your data and use your tools, the question goes beyond choosing a model. It becomes a matter of architecture, identity and operations.

What the agent doesWhat that requires
It reads your dataPermissions enforced at the source and separated indexes, so it only finds what the user can already see.
It acts through an accountA dedicated identity, least privilege and human approval for sensitive actions.
It reads emails and web pagesTreating that content as untrusted, because it can carry malicious instructions (prompt injection).
It consumes resourcesRealistic sizing of cost, latency, GPU capacity and quotas, with a recovery plan for outages.
It can get things wrongLogs, evaluations and the ability to roll back, so you can see what happened and fix it.

Guidance and agent operations

Going live is only the beginning.

An agent changes along with your data, your models and how people use it. I guide you through AI adoption and run your agents over time, with your teams or on their behalf, depending on the size of your company.

AI guidance

Helping your leaders and teams decide, adopt and stay self-sufficient.

Direct
vision and priorities shared with leadership
Frame
usage rules and AI governance
Train
building your teams' skills
Advise
technical and vendor choices

Agent operations

Knowing at any time what your agents are doing, what they cost and whether they remain reliable.

Monitor
activity, errors and sensitive actions
Evaluate
quality and drift after every change
Control
costs, quotas and capacity
Govern
agent inventory, permissions and human approvals
Improve
incidents, model and prompt versions

Services

From the first use case to daily operations.

  1. 01

    Agentic scoping

    Find the use cases worth pursuing, assess your prerequisites and decide what to build, buy or drop.

    • Use cases with measurable value
    • Data, identity and systems review
    • Cloud, hybrid or local choice
    • Prioritised roadmap
  2. 02

    Agent architecture and integration

    Connect agents to your tools and systems with an architecture your teams can operate.

    • Single and multi-agent orchestration
    • Connectors and MCP servers
    • RAG on your internal sources
    • Human oversight and observability
  3. 03

    Local and sovereign AI

    Run your models inside your own perimeter when confidentiality, regulation or cost call for it.

    • Open-weight models on your servers
    • GPU sizing and inference
    • On-prem deployment, isolated if needed
    • Hybrid routing between local and cloud
  4. 04

    Agent security

    Define what agents can see and do before they touch production.

    • Agent identities and least privilege
    • Protection against prompt injection
    • Sandboxing and action approval
    • Traceability to support compliance (FADP, GDPR, AI Act)
  5. 05

    Messaging and AI

    Put agents to work on messaging without turning it into a new way in for attackers.

    • Triage, search and reply agents
    • Microsoft 365 and Exchange Online
    • SPF, DKIM, DMARC and MTA-STS
    • Control of inbound content
  6. 06

    On-prem and cloud foundations

    Bring the systems your agents rely on up to standard, so agents don't weaken what already works.

    • On-prem and hybrid infrastructure
    • Azure, AWS and Microsoft 365
    • Identity and access
    • DNS, network, backup and continuity

Architecture

Where should your agent run?

There is no universal answer. The right choice depends on how sensitive your data is, your volumes and your teams. Pick a scenario to see what changes.

Keep your data in-house and send to the cloud only what can go there.

A good fit when
Data of varying sensitivity, an occasional need for very powerful models.
Watch out for
Clear routing rules, filtering of what leaves and more components to operate.

Your perimeter : Agents and interface, Orchestration and tools, Models, Data and indexes, Identities and permissions, Logs and audit

Cloud : Agents and interface, Models

Background

25 years of infrastructure, one new layer.

Agentic AI is the newest layer of my career. All the others are what keep it standing in production.

One senior point of contact

My career was built in demanding, international environments where availability and confidentiality are not negotiable. You speak directly with the person guiding you, from scoping to operations, and you get clear explanations and prioritised recommendations.

  1. Agentic and local AI

    Agents, RAG and local models, designed for operations from the start.

    • Agents
    • MCP
    • RAG
    • Local models
    • GPU

  2. Security

    A pragmatic approach to access, data protection and incidents.

    • IAM
    • MFA
    • Network segmentation
    • Backup
    • Incident management

  3. Messaging

    Enterprise messaging, collaboration and email authentication.

    • Exchange Online
    • Microsoft 365
    • SPF · DKIM · DMARC
    • MTA-STS
    • DNS

  4. Cloud

    Integration, security and governance of cloud and SaaS platforms.

    • Azure
    • AWS
    • Microsoft 365
    • SaaS integrations

  5. On-prem infrastructure

    The foundations: servers, networks and critical environments with high availability demands.

    • Servers
    • Network
    • VPN and firewall
    • Telephony and VoIP
    • Continuity

Method

From the first idea to day-to-day operations.

  1. 01

    Scope

    Understand your goals, data and constraints, then choose the right first use case.

  2. 02

    Prove

    Run a pilot on a real case, with success criteria set in advance.

  3. 03

    Secure and deploy

    Harden access, set up monitoring and move to production in stages.

  4. 04

    Operate

    Track quality, costs and incidents, with your teams or on their behalf, then adjust your agents over time.

  5. 05

    Hand over

    Document and train, so your teams stay in control of their agents.

Offers

Three ways to get started.

  • One-off

    Agentic assessment

    Know where you stand and where to start.

    • Use case review
    • State of data, identity and security
    • Cloud, hybrid or local recommendation
    • Prioritised roadmap

  • Project

    Pilot in production

    Put a first agent into service on a real case.

    • Architecture and integration
    • Security and human approval
    • Monitoring and evaluation
    • Documentation and handover

  • Ongoing

    Guidance and operations

    A senior expert to run your agents with your teams or on their behalf.

    • Regular reviews of your agents
    • Tracking quality, costs and incidents
    • Security and review of new architectures
    • Team skills development

Scope and price agreed together after a first conversation.

Let's talk about your project.

Describe your context in a few lines, and what you would like to hand over to agents. No specification needed.

Working in Switzerland.

contact@arntechcloud.ch