Kotlin Desktop Engineer – AI Inference Tooling (Senior/Staff)

VirtusLab is a leading European software consulting and engineering company, home to over 350 EU-based professionals. Our mission is to craft clean code and practical solutions with precision and purpose. We foster a dynamic culture rooted in strong engineering, a sense of ownership, and transparency, empowering our team. As part of the expanding VirtusLab Group, we offer a compelling environment for those seeking to make a substantial impact in the software industry within a forward-thinking organization.

About the role

We’re looking for an engineer to join a small, startup-like team building a desktop tool for local AI inference. You’ll work with Kotlin and Compose Desktop to create an application that deploys, configures and manages AI models running on local devices – think NVIDIA Spark, Claude Code, and Codex connected to locally hosted models.
Your daily work will include building the UI in Compose Desktop, integrating with backend REST APIs for model deployment, and setting up agent configurations through TOML/JSON/YAML files. You’ll also work on LLM proxy routing and create installers for Windows, macOS and Linux. The team already has an internal MVP and talks directly with potential users. We’re on the road to making it open source. If you want to shape a developer tool from the early stages, this is it.

Project

Local Inference Platform

Project Scope

We are building a desktop application for deploying and configuring local inference on local devices (e.g. NVIDIA Spark / DGX Spark), managing model lifecycle (start/stop), and connecting coding agents like Claude Code and Codex to locally hosted models. Currently an internal MVP, on the road to open source.

Tech Stack

Kotlin, Compose Desktop (with JetPack Compose roots), kotlinx.coroutines Backend integration: REST APIs for model deployment management Agent layer: configuration files in toml/json/yaml, LLM proxy routing from agents to BE Packaging: native installers — msi (Windows), deb (Linux), dmg (macOS) Nice-to-have tooling: mitmproxy for HTTP debugging, LiteLLM/OpenRouter-style proxy layers.

Challenges

Building a desktop control plane for local AI agents that abstracts away the messy parts of running LLMs on heterogeneous hardware — model lifecycle, proxy routing, agent configuration, and OS-specific quirks (path separators, signing, packaging) — while keeping the system extensible enough to evolve toward open source. Visualising model statistics and system load on top of a backend that actually does the deployment.

Team

Small, startup-like team on the JetBrains side — no bureaucracy, direct contact with potential users, fast iteration. Code reviews as part of the process.

What we expect in general

• Strong Kotlin skills and genuine interest in Compose Desktop as a UI framework

• Understanding of asynchronous and concurrent programming with Kotlin Coroutines

• Awareness of cross-platform differences (file paths, OS-specific packaging)

• Engineering pragmatism: cutting complexity while keeping the system extensible

• System-level thinking: understanding how UI, backend APIs, agent configs and proxy layers fit together

• Hands-on experience using LLMs daily, with the judgement to know when they help and when they don’t

• Fluency in English, with good communication skills for a remote-first team

• Self-motivation and the ability to take full ownership of features end-to-end

• Experience with different LLM agents and their configurations (nice to have)

• Experience with mitmproxy or similar HTTP traffic inspection tools (nice to have)

• Experience with LLM proxy layers such as LiteLLM or Openrouter (nice to have)

• Experience creating and signing OS installers (MSI, DEB, DMG – nice to have) 

A few perks of being with us

• Building tech community

• Flexible hybrid work model

• Home office reimbursement

• Language lessons

• MyBenefit points

• Private healthcare

• Training Package

• Virtusity / in-house training

And a lot more!