AI and cloud systems that reach production
AI Advisory · Solutions Architect · Fractional CTO
I help leadership teams work out which AI use cases are actually worth funding, then build the first one — on infrastructure that holds up under audit. Twenty years of cloud architecture, with the same person writing the roadmap and the code.
See What You Get- Assessment · 2–3 weeks · from €10,000
- Fee credited against the build
- Your cloud account, EU region
Who I Have Worked With
Past and current clients, across public sector, automotive, travel, hospitality, and financial services.
- European Commission
- Volvo USA
- TUI Group
- Ricoh
- Talkdesk
- Metro (Hospitality Digital)
- Q.ai (Forbes)
What You Get
Fixed scope, fixed price, fixed duration. Start anywhere on the ladder and stop after any step.
AI Readiness & Opportunity Assessment
The engagement most clients start with. I interview the people who do the work, look at your data and infrastructure as they actually are, and come back with a sequenced plan you can take to both your board and your auditors.
- Prioritized use-case portfolio, with estimated cost, effort, and payback per case
- Data and infrastructure readiness gaps, named and sequenced rather than listed
- AI Act classification of each candidate use case and the obligations that follow
- A build-versus-buy call for every shortlisted case, with the reasoning
- Reference architecture for the first use case, costed at your expected volume
- A 12-month roadmap built around decision points, not a wish list
Fee credited in full against a Proof of Value build booked within 90 days.
Book an intro callAI Opportunity Workshop
A structured working session with your leadership and operations leads. Often the cheapest way to find out that the thing everyone is excited about is not the thing worth doing first.
- Long list of candidate use cases, scored on value and feasibility
- Written summary of the three worth pursuing, and why the rest are not
- A clear read on whether a full assessment is warranted
Remote or on-site anywhere in the EU.
Proof of Value Build
One use case, built and running against your real data. This is where a strategy engagement normally ends and where mine keeps going.
- Deployed in your own cloud account, EU region, under your IAM
- Evaluation harness, so answer quality is measured rather than asserted
- Running cost measured at real volume, not estimated in a spreadsheet
- Handover documentation and a working session with the team who will own it
Fractional AI Leadership
A senior technical voice in the room for companies that need AI decisions made well but cannot justify a full-time hire to make them.
- Standing time with your leadership team, weekly or fortnightly
- Architecture and vendor decisions made, documented, and defensible later
- Review of work delivered by internal teams and outside suppliers
- Hiring input for AI and platform roles
Three-month minimum, then rolling monthly with 30 days' notice.
Team Training
Separate tracks for leadership and for engineers, built around your stack and your real use cases rather than a generic curriculum.
- Delivered by an AWS Authorized Instructor
- Leadership track covering what to fund, what to refuse, and what to ask for
- Engineering track covering RAG, evaluation, cost control, and failure modes
- Exercises and material your team keeps afterwards
Prices are per engagement, excluding VAT and travel, and scale with company size and scope. Fixed price means that if the work runs long, that is my problem and not your budget's.
Also Available
Cloud Architecture
Architecture design across AWS, GCP, and Azure — serverless, microservices, Kubernetes (EKS, GKE), event-driven systems with Kafka and SQS/SNS, and API management with Apigee and WSO2. I apply Domain-Driven Design to decompose legacy monoliths into maintainable, independently deployable services.
Cloud Migration
Full-cycle migrations from on-prem to AWS and GCP. Past work includes a Magento 2 platform to Kubernetes on EKS, healthcare systems, and computer vision workloads. I handle architecture, risk planning, FinOps (rightsizing, reserved instances, cost dashboards), and delivery.
DevOps as a Service
Infrastructure as code with Terraform and Ansible, containers with Docker and Kubernetes, monitoring with Datadog and Prometheus. Security covered: AWS WAF, Shield, IAM, and Secrets Manager. CI/CD across GitLab, GitHub Actions, BitBucket, and Azure DevOps. For teams that need DevOps capacity without a full-time hire.
Serverless Architecture
Serverless system design and implementation using AWS Lambda, API Gateway, Step Functions, and the Serverless Framework. I have built serverless backends for a healthcare startup (acquired by Carbon Health), a loyalty card platform, and a litigation support system.
Selected Work
LLM Document Search Without the Data Leaving
A legal technology client needed large volumes of sensitive documents made searchable and summarizable, with a hard constraint that nothing could be sent to an external service. Built a RAG knowledge base on Amazon Bedrock inside their own AWS environment. Lawyers ask questions in plain language and get sourced, referenced answers rather than a list of search hits.
Clinical-Grade Platform, No Servers to Manage
A healthcare startup needed a backend that could handle unpredictable clinical load and meet reliability requirements without a team to run infrastructure. Designed a fully serverless architecture on AWS using API Gateway, Lambda, the Serverless Framework, and Kafka. The company was subsequently acquired by Carbon Health.
Cloud Foundation for a Global Travel Group
One of the world's largest travel companies needed a cloud native foundation built from scratch that was production-ready and auditable from day one. Delivered the full stack on AWS with Terraform for infrastructure as code, GitLab CI/CD, Apigee for API management, and Datadog for monitoring.
E-Commerce Platform Migration to Kubernetes
A retail client running a Magento 2 platform on ageing infrastructure needed a setup that could scale and was cheaper to run. Migrated to Kubernetes on AWS EKS using Helm for deployments, RDS for the database layer, and S3 for storage. The team went from manual, error-prone deployments to a fully repeatable, environment-consistent process.
One person. No handover to juniors.
I’m Ivan Prlić, founder of Prlić Consulting. Twenty years in cloud architecture, most recently building AI systems that had to survive a compliance review before they were allowed anywhere near production.
You deal with me directly throughout. The person on the scoping call is the person writing the roadmap and the person writing the code, which is the main practical difference between this and hiring a firm.
Prlić Consulting d.o.o. is registered in Zagreb, Croatia. I work remotely across EU and US timezones and travel for on-site work in the EU.
Amazon Bedrock · SageMaker · Vertex AI · Azure OpenAI · AWS · GCP · Azure · Kubernetes · Terraform
English (C2, Cambridge Certified) · Croatian (Native)
Before You Ask
The questions that decide these engagements are rarely about capability. These are the ones procurement and legal raise.
Where does our data actually live?
In your own cloud account, in an EU region, under your own IAM. I build inside your perimeter rather than in mine, so there is no vendor environment holding your data and nothing to unwind if we stop working together. On the legal document search system I built on Amazon Bedrock, no document ever left the client’s own AWS account, which is what made it possible for their legal team to sign off.
Have you worked under formal compliance requirements before?
Yes. I designed the cloud architecture for Q.ai, an AI-powered investment platform, under financial-grade requirements including SOC 2 controls, encryption, and audit logging. I have built healthcare systems with clinical reliability requirements, delivered for the European Commission, and worked on enterprise systems at Volvo USA and TUI Group where security and governance frameworks were set by the client. I am used to architecture decisions having to be explainable to an auditor.
Who actually does the work?
I do, all of it. The person on the intro call is the person writing the roadmap and the person writing the code. There is no bench, no handover to a delivery team, and no junior learning on your budget. The trade-off is honest: I am one person, so if you need ten, I am the wrong answer.
What does the AI Act actually require of us right now?
Less than most vendors imply. Obligations for general-purpose AI models have applied since August 2025. The deadlines for high-risk systems were pushed back by the Digital Omnibus to December 2027 for Annex III and August 2028 for Annex I. So this is a preparation window rather than an emergency. The assessment classifies each of your candidate use cases and tells you which obligations follow, so you can plan against real dates instead of a sales pitch.
Can you work alongside our internal team or an existing supplier?
Usually that is the arrangement. At Metro’s Hospitality Digital I led architecture across several existing development teams, at Ricoh I coached an in-house team, and at TUI Group I guided their developers on cloud-native patterns rather than replacing them. Part of the fractional retainer is reviewing work delivered by internal teams and outside suppliers.
What happens if the assessment concludes we should not do this?
Then that is the deliverable, and it is a good outcome. Finding out in two weeks that a use case does not pay off is worth considerably more than finding out after a six-figure build. I would rather write that than take on delivery work I do not think will hold up.
Are you an EU entity, and how does invoicing work?
Prlić Consulting d.o.o. is a Croatian limited company, registered in Zagreb since November 2017, OIB 84570551730. EU entity, EU invoicing, EU data residency. I work remotely across EU and US timezones and travel for on-site work anywhere in the EU.
Start with a call
A short call, no charge, no deck. We work out whether there is a real opportunity here and whether I am the right person for it. If the answer is no, you will hear it on the call.