Services
Architecture support for backend, data and AI systems
I help founders and engineering leaders make the decisions that are expensive to undo, and get them done without a rewrite.
Architecture review
A focused look at whether your system fits what the business needs next, before a decision that is hard to undo.
How it runs
The right architecture depends on the right requirements, so I start with the business: your goals, your customers and what the system must do for them. Then I read the code, the architecture, the data model, the cloud setup and how you deliver, and I talk to the people who built it. We agree the scope and the price after the intro call.
When teams bring me in
- When the business changes course: a new market, a new product, or many more customers
- Before a rewrite, or a move from one codebase to many services
- When the cloud bill grows faster than the users
- When the database or the data pipelines start to hold the product back
- Before AI coding agents get access to the codebase
- When the engineers disagree about the architecture
What you get
- The business requirements behind the system, written down, so the business and the engineers agree on what it must do
- A written report: how well the system fits those requirements, what is solid, what is risky, and what to change first, with the trade-offs and a rough effort for each
- A short decision log for the choices that matter
- A walkthrough with your team
Focused sprint
One milestone, done with your engineers, then handed over.
How it runs
Tied to one milestone the business needs, not open-ended hours. I work in your code where that helps, in Java, Scala or Python, and pair with your engineers so the knowledge stays.
Examples
- Split a monolith into services step by step, without downtime
- Design the database and data model for what the product needs next
- Move from overwritten rows to events: an event store, projections, separate reads and writes
- Make data pipelines reliable: backpressure, retries and replay
- Take a machine learning model from an experiment to a production service, with monitoring and rollback
- Cut the cloud bill with right-sized, simpler infrastructure
- Roll out AI coding agents with rules, review and tests
- Make a web app findable by Google and AI assistants
What you get
- Working software, or a design your team can build
- The decisions written down
- A team that can carry on without me
Ongoing second opinion
A few days a month for the decisions that keep coming up.
How it runs
A set number of days a month, agreed together and easy to change.
What it covers
- Design reviews before the code is written
- A check on new tools, vendors and AI agents before they spread
- An outside view when the team can't agree
Areas I work in
- Java and Scala backends: with ZIO, Spring or whatever fits the team, reliable in production and easy to change
- Databases and data platforms: PostgreSQL, Cassandra, MongoDB, and pipelines that stay clean
- Event-driven systems: Kafka, event sourcing and CQRS
- Modernization and digital transformation: step by step, from old systems and manual processes to software that is easy to change
- Cloud: AWS, Azure and European providers, with an eye on cost and where data lives
- AI in production: data quality, latency, evaluation and serving
- AI coding agents with guardrails: rules, review and security
- Search and AI visibility: websites that Google and AI assistants can read, with pages crawlers can use, structured data and access for AI crawlers
How it works
How a project runs
The goal is not more complexity. It is a system your team can run, improve and afford.
A 30-minute intro call
Your system, your constraints and your goals. You leave with a clearer next step.
A look at what is there
The architecture, how you deliver, the data model, the cloud setup and the risks.
A plan
The target architecture, the decisions behind it, and a roadmap your team can follow.
Hands-on or guidance
I build key parts with your engineers, or guide them while they do.