Muhamed Isabegovic Get in touch

About

I have spent 10+ years in the industry taking ownership of products and the teams that build them, from deciding what to build to running it in production, anywhere in the stack. I have built trading systems in C++, SaaS products in Rails and AI features in hiring software.

Most recently I built AI screening features for Teamtailor’s Co-pilot team. Before that I was tech lead for a team of 10 engineers at Recrubo, through its acquisition by Carv.

On Usput.ba, my own travel guide to Bosnia and Herzegovina, I work as an AI engineer. Descriptions, audio tours and itineraries are generated from real place data rather than from the model’s memory, and it is where I try new AI features on real users first.

Lately I also work one layer up, with agents writing the code. I build open-source tools that keep that reliable. pi-brain holds the specs and decisions the code is regenerated from, and runs this site. enola checks code against architecture rules; I am a core contributor and maintain four gems that bring it to Ruby and Rails teams.

I also help companies become AI-native, in how they build and in what they build; see Services.

Career Overview

  • (2024 - present) Founder, own company (Isabegovic Muhamed), Vienna: alongside Meisterlabs in 2024, my main work since 2025
  • (2021 - 2024) Senior Software Engineer @ Meisterlabs
  • (2019 - 2021) Lead Software Engineer @ Experfy
  • (2018 - 2019) Software Engineer @ Marvelsoft
  • (2016 - 2017) Junior Software Engineer @ Experfy

Education and languages

  • Master’s studies in Software Engineering and Internet Computing, and Business Informatics, at TU Wien, part-time alongside work, since 2022
  • BSc in Information Technology, IPI Akademija Tuzla
  • Bosnian (native), English (full professional), German (limited working), Spanish (elementary)

Specialties

  • AI product engineering (RAG, LLM pipelines, evals, structured outputs)
  • AI agents and MCP, with guardrails for agent-written code
  • Spec-driven development
  • Software architecture and design
  • Technical leadership and engineering management
  • Community building and conference organising

Open source

  • pi-brain: a cloneable, agent-maintained knowledge base. This site runs on a clone of it.
  • Core contributor to enola: architectural regression testing for AI-assisted development. My contributions include the constraints program and its verdict vocabulary, plan and constraints mine, the fact-provider seam and its providers, the history store behind blame and diff, declared intent compiling into the graph, Ember support, the Rails extraction work (dead-methods, query-loops), the Ruby DSL for writing laws as sentences, and the verdict writers that put findings where CI reads them (see the project’s acknowledgements).
  • Maintainer of four enola gems:
    • enola: pure-Ruby wrapper over the released enola binary, with the Ruby providers (Prism, Rubydex) on by default.
    • enola-rb: the Rails layer. A generator plus enola:init, enola:snapshot, and enola:check; a fresh rails new goes from nothing to a caught architecture breach in seconds.
    • enola-guides: the content layer. Guides, a catalogue of architectural laws, worked examples, agent skills, and CI tooling.
    • munola: my own channel on top of enola, with a built-in recipe catalogue bound by detection.

Tech stack

  • What I build: RAG, LLM pipelines, LLM evaluation, Structured outputs, Spec-driven development, Agent guardrails, Harness extensions, Prompt-injection screening
  • AI and agents: Claude, OpenAI, Gemini, RubyLLM, MCP, Langfuse, pgvector
  • Languages: Ruby, Go, Rust, TypeScript, Python, JavaScript, C++
  • Frameworks and data: Ruby on Rails, Sidekiq, Ember.js, React, Vue.js, Hotwire, Node.js, Redis, PostgreSQL, MySQL, DuckDB
  • Infrastructure: Docker, Kubernetes, Helm, Google Cloud, CI/CD, Kamal

Things I love

  • My family
  • Committing to a good cause
  • Good people & good stories
  • Movies (since I was very little)
  • Coffee
  • Ruby (Community)