AI agents that work inside your company.
An agent your team talks to every day, workflows that take the routine work, and the private platform underneath. All of it on your infrastructure.
- Who you talk to
- The architect. Same person on the first call and the incident bridge.
- Where it runs
- Your cloud account, your tenant, your rules. Keys, memory and logs stay with you.
- How it is priced
- Fixed scope or retained advisory. Price and deliverables in writing before anything starts.
The situation
The situation
-
Your team already uses AI. On personal accounts, with company data.
-
The pilots impressed everyone and shipped nothing.
-
Your next audit will ask where the data went.
This is not a model problem. It is an architecture problem.
Services
Six ways to work together.
From a first assessment to a private AI platform. Each one ends with your team owning the result.
-
AI Readiness Assessment
Where AI pays off in your company, where it will hurt, and what to build first. Two to three weeks, one written plan you can execute with me or without me.
-
Agent Deployment
Hermes Agent, or the open-source runtime that fits, installed on your infrastructure and connected to your tools. An AI colleague for your team, under your rules.
-
Agent Workflows
Ticket triage, document intake, reporting, ops routines. Repetitive work handed to agents, with a person approving where it counts. Measured, not promised.
-
Private AI Platform
Models, gateway and retrieval on your cloud. Your documents answer questions with the access rules they already have. Every call logged, every cost traced to a team.
-
AI Security and Governance
Agents get identities, permissions, logs and limits like any other user. LGPD, GDPR, ISO 42001 and NIST AI RMF turned into controls that run, not binders that sit.
-
Cloud and Security Foundations
The platform underneath the AI. Landing zones, identity, CI/CD with security gates, compliance as code, and rescue for estates that grew by accident.
Compare all services: problems, outcomes, deliverables and the ideal client for each
What changes
Four things you notice after the work is done.
-
Your team gets hours back.
Drafting, research, triage, reporting. The routine work moves to agents. People keep the decisions and the approvals.
-
AI stops being a shadow risk.
One agent your company runs replaces the personal accounts your data was leaking into. Logs, permissions and limits included.
-
Every token makes sense.
Costs trace to teams and workflows. Model choice is a setting, not a rewrite. No more "what is this subscription for?"
-
The board question has an answer.
Where does the data go, who approved what, which controls run. The answer is a query, not a meeting.
How it works
Talk. Scope. Build. Stay.
Four steps, no surprises. You see the price and the deliverables before anything starts.
-
Talk.
30 minutes. I learn your team, your tools, your constraints and what AI must never touch. No templates. No questionnaires.
-
Scope.
Clear proposal. Fixed deliverables. Price adjusted to your budget. You know exactly what you are paying for before we start.
-
Build.
I deploy alongside your team, one tool and one workflow at a time. Everything as code, everything documented. When we are done, your team owns it.
-
Stay.
If you want. Ongoing advisory for teams that want a trusted architect available. There when you need me.
Selected work
Problems I have taken from broken to boring.
Client names stay private. The structure never changes: challenge, approach, deliverables, measured outcome.
-
0 client documents in personal AI accounts after rollout
Deployed a Hermes Agent on the firm's own cloud tenant, connected to mail, calendar and documents with per-person access, replacing personal chatbot accounts and taking recurring drafting and research work off the partners' desks.
-
-71% first-response time on routine categories
Built an agent workflow that classifies, enriches and drafts answers for inbound support tickets, resolves the routine ones with one-click approval and escalates the rest with full context. First-response time fell while a person stayed on every risky reply.
-
-38% monthly cloud spend after decommissioning
Consolidated eleven hand-built environments into a single orchestrated Kubernetes platform with GitOps delivery, cutting the cloud bill and the incident rate at the same time.
Who you work with
Lucas
Founder, AI and cloud architect. The same person on the first call, the design review and the incident bridge.
I run open-source AI agents on my own infrastructure every day, and I build and benchmark agent tooling. I know what breaks, what leaks and what actually saves time, because I have measured it.
Before that I spent years as a cloud and security architect. I've been the engineer debugging production at 1 am (true story). I've built SOC environments and designed identity architectures across Azure, AWS and GCP, at companies with 15 people and companies with 9,000. That is the foundation the agents run on.
ataides. exists because I got tired of watching good teams lose months to bad architecture. AI is no exception.
Next step
Let's build yours.
Tell me what is slowing your team down. I answer within one business day, and the first 30 minutes are on me.