Jakub Havelka
Portrait of Jakub Havelka, senior software engineer based in Bratislava

Senior Java/Spring engineer shipping AI features into production codebases.

I help European SaaS and fintech teams add RAG, LLM integrations, and AI-powered workflows to the JVM and Angular systems they already run — without rewriting what works.

See recent work

Credibility

12+ years

in professional software engineering, including 4 years building order management infrastructure for a German bank serving billions in daily securities transactions.

Production AI

running on cravingtoolkit.com — RAG and content generation in a privacy-sensitive domain.

Full stack to full pipeline

Java, Spring Boot, Angular, Kafka, plus LangChain, LangGraph, vector DBs, and modern LLM APIs.

What I do

I'm Jakub Havelka, a senior software engineer based in Bratislava. For four years I've built and operated production systems at dwpbank (Deutsche WertpapierService Bank) — the kind of Spring Boot, Kafka, and Angular stack that quietly powers the European financial system. Now I help teams with similar codebases ship AI features that are actually production-grade: properly observed, cost-controlled, and built to survive contact with real users.

Why this matters

Most AI freelancers can prototype on a laptop. Fewer can integrate into a large Java codebase with auth, audit logging, rate limits, and a CI/CD pipeline that has to keep working tomorrow. That's the gap I fill.

Recent work

cravingtoolkit.com — RAG and AI content pipeline

A recovery-focused content platform with an automated SEO publishing pipeline and a RAG-powered Q&A layer. Demonstrates LLM safety considerations in a YMYL (Your Money or Your Life) domain, content generation guardrails, and serverless retrieval architecture.

Stack: Python, OpenAI and Anthropic APIs, vector embeddings, automated content workflows.

Outcome: Live AI features serving real users; documented architecture, eval harness, and editorial guardrails.

Read the case study →

Order Management Systems at dwpbank

Four years of fullstack development on the Kundenanbindung module of dwpbank's Order Management System. Production Spring Boot, Apache Kafka, JPA, Angular, and AWS — the kind of infrastructure that processes German retail and institutional securities orders without room for downtime.

Stack: Java/Spring Boot, Apache Kafka, JPA, Angular, AWS, GitLab CI/CD.

Outcome: Production systems contributing to a platform that custodies hundreds of billions in client assets.

Read the case study →

Multi-agent automation infrastructure

Self-hosted AI agent infrastructure running multiple LLM-powered agents on a managed cloud node, integrated with messaging platforms and routing across multiple model providers. A practical study in production agent reliability: fallbacks, cost management, and context window strategy across provider migrations.

Stack: Linux infrastructure (Hetzner Cloud), multiple LLM providers (Anthropic, MiniMax, Gemini), Telegram integration, agent frameworks.

Outcome: Working multi-agent system with documented architecture, migration path, and cost discipline.

Read the case study →

How I work

Scoped engagements

Two- to twelve-week projects with clear deliverables. No open-ended retainers, no surprise scope creep — written scope before kickoff, weekly demos, and a clear exit at every milestone.

Embedded into your codebase

I don't drop a Jupyter notebook and walk away. AI features ship into your existing Spring Boot, Angular, or Node codebase with the same observability, auth, and CI/CD discipline as the rest of your platform.

Honest about what I don't do

I'm a production engineer with deep AI integration experience — not a research scientist. If your project needs novel architecture work, fine-tuning from scratch, or distributed training, I'll tell you and refer you to someone who does that for a living.

Stack

Backend / production
Java 17+, Spring Boot 3.x, Kafka, JPA/Hibernate, PostgreSQL, AWS, GitLab CI/CD, Docker
Frontend
Angular, TypeScript, basic React
AI / LLM
Anthropic Claude, OpenAI, Gemini, open-source models via Hugging Face; LangChain, LangGraph; vector DBs (Pgvector, Pinecone, Qdrant); RAG pipelines; agent orchestration; structured output and function calling; production eval and observability
Infrastructure for AI workloads
Hetzner Cloud, AWS, basic Kubernetes, serverless functions

Who I work with

Good fit

  • European SaaS or fintech teams with a working product and a Java/Spring or Angular codebase
  • Teams that want a senior engineer who can ship AI features end-to-end, not a contractor who needs hand-holding
  • Projects between 2 and 12 weeks with a clear definition of done
  • Companies that care about evals, observability, and not lighting OpenAI credits on fire

Not a fit

  • "Build me an AI app" with no spec — I'll help you write the spec, but I won't pretend it doesn't matter
  • Pure research, pre-training, or model architecture work
  • Web3 / crypto projects (the security risk profile in this niche is too high right now)

Let's talk

Tell me what you're building and I'll tell you honestly whether I can help.

Email: hello@jakubhavelka.dev

Based in Bratislava, Slovakia (CET, UTC+1/+2). Working with clients remotely across Europe.