◆ How I Work
Built precisely. Delivered fast.
Six principles that hold across every engagement mode: training your team, building production infrastructure, embedding as a senior operator, or supporting an outsourcing firm. The shape changes. The standards do not.
◆ Principles
Focused engagement, not divided attention.
Full attention · No divided focus
I work with a small number of clients at any time. Not because capacity is constrained, but because quality requires focus. When I take on a project, the client gets full engineering attention. Architecture decisions, production deployments, technical communication: these do not get delegated or diluted.
You work with me. Not around me.
Direct engagement, founder to engineer
There is no team of junior developers behind me doing the implementation while I review. No PM layer translating your requirements. No agency overhead. The person on the proposal is the person writing the code. Every architecture decision, every deployment, every production incident: that is me.
Production on day one. Not after a design sprint.
Week 1: working system. Not a wireframe.
I do not build demos. I do not run two-week discovery processes before writing a line of code. By the end of the first week of an engagement, there is working software. The architecture decisions that make a system maintainable at scale are made before the first commit, not retrofitted later.
Agentic development as a competitive advantage.
18 months of agentic workflow refinement
I use Claude Code and Cursor as my primary development environment, not as productivity supplements. The practical result: I operate with the leverage of a small team. Problems that would require three engineers and a sprint cycle to solve take days. This is not a productivity claim. It is a structural change in what a solo operator can build.
Async by default. Meetings when they matter.
Serbia · GMT+2 · North America clients welcome
Most decisions can be made asynchronously. I document clearly, communicate proactively, and answer questions same-day. Weekly video demos replace status meetings. The Jira Olympics are not part of the engagement. Time difference between Serbia and North America works in our favour: your morning has a full day of work waiting.
No ongoing lock-in. You own everything.
Full IP transfer · Clean handoff documentation
All code, all documentation, all infrastructure configuration is yours. When the engagement ends, the system keeps running. I write handoff documentation as a first-class deliverable, not an afterthought. The goal is a system your team can maintain and extend without me, and a founder who calls me back for the next problem because the first engagement was worth it.
◆ What I Build
LLM system architecture
- ◆Multi-agent orchestration design and implementation
- ◆RAG pipeline architecture (retrieval strategy, chunking, evaluation)
- ◆Production LLM feature development end-to-end
- ◆Prompt engineering and evaluation frameworks
- ◆Context window management and cost optimisation
MLOps & infrastructure
- ◆Model deployment pipelines (training → staging → production)
- ◆Monitoring and observability for LLM systems
- ◆Retraining automation and model versioning
- ◆Cost tracking and inference optimisation
- ◆Edge deployment and serverless ML
Full-stack AI product
- ◆Next.js / FastAPI full-stack builds with AI core
- ◆Real-time AI features (streaming, low-latency inference)
- ◆Compliance automation (SOC 2, ISO 27001, HIPAA, PCI-DSS)
- ◆Agentic workflows for complex business processes
- ◆Third-party integrations and data pipelines
◆ Engagement Structure
For build and embedded engagements. Training and partner arrangements are scoped separately.
Week 1
Alignment
Architecture review, tech stack confirmation, risk identification, first working component delivered by Friday.
Weeks 2–4
Core build
Primary system components built, tested, and integrated. Weekly Friday demos. You see real software, not status updates.
Month 2
Production
First production deployment. Monitoring, alerting, and observability in place. Real users on the system.
Ongoing
Iteration
Weekly demos, async-first communication, continuous shipping. Every engagement runs until the system is in your team's hands.
◆ Who I work with
The engagement shape depends on what you need. The client type does not.
I work with startups building AI products from scratch, enterprises trying to make their engineering teams faster, outsourcing firms who need a senior technical layer behind their accounts, and companies who need an experienced AI operator inside the team for a defined window. The four engagement modes exist precisely because the need looks different depending on where you are.
◆ What every engagement has in common
- ◆A technical stakeholder who can evaluate the work
- ◆A defined problem with a definable outcome
- ◆Direct communication, both ways
- ◆A mandate to build or improve, not just explore
◆ Genuinely not a fit
- ◆Companies that want AI as a slide, not a capability
- ◆Scope-less projects that resist definition after two conversations
- ◆Clients who need a vendor to agree with everything
- ◆Pure legacy maintenance with no AI transformation angle
If the brief matches, I would like to hear from you.