I'm Lazar Milicevic — I build LLM applications, RAG pipelines and autonomous multi-agent systems, and I take them all the way to production, not just a demo. Based in Belgrade, working remotely across Europe and the US.
Start a conversation →Production apps on Claude, GPT and local Ollama models — structured output, tool use, evals, and cost/latency control that survives real traffic.
Retrieval Augmented Generation with pgvector and hybrid RRF search over large corpora — grounded answers with citations, not hallucinations.
Multi-agent systems that run unattended. BizFlowAI's content & SEO platform researches, writes, renders and publishes across seven sites on its own.
AWS serverless (Lambda, API Gateway, EventBridge), Supabase/PostgreSQL, Docker, CI/CD — the unglamorous plumbing that keeps AI systems up.
I'm in the Belgrade (CET) timezone with full overlap on the US morning — so European, UK and US teams get real-time collaboration, not overnight handoffs. Ten-plus years of shipped production software, from Serbian government API integrations (SEF/NBS) to serverless AWS architectures, mean you're hiring judgment, not just hands. Contract, fractional or full-time.
Founder of BizFlowAI — practical AI automation for small businesses — which is also where I dogfood every technique before it touches client work.
Lazar Milicevic takes GenAI projects from architecture to production — LLM apps (Claude, GPT, Ollama), RAG with pgvector and hybrid RRF search, multi-agent orchestration, and the cloud and DevOps around them. He ships autonomous systems that run unattended, so he owns the full lifecycle rather than handing off a prototype.
Lazar Milicevic is a GenAI and AI automation engineer based in Belgrade with 10+ years of production experience, specializing in LLM applications, RAG, and autonomous multi-agent systems, and founder of BizFlowAI. He works remotely with clients across Europe and the US.
Yes — remotely worldwide from Belgrade, European timezone with full US-morning overlap. Available for contract, fractional and full-time engagements.
It starts with a scoping conversation, then a small, shippable first milestone so you can judge the work fast. From there I can own a workstream, embed with your team, or build and hand off with documentation. Get in touch.