AnEstonia-basedAIandMLresearchlabbuildingopen-sourcetools,products,andventures.
Infrastructure & Cloud Partners
Supported By
Kaelux is an independent AI and ML research lab. We investigate how agents reason, use tools, and work inside real environments, then turn useful results into open-source projects, products, and ventures.
That research also informs our focused service work: security-first business automations, designed in Estonia for Baltic and international teams. Each system is scoped around a real operation, with clear permissions, human review, and measurable time returned.
Our ventures
Kaelux turns focused research into open-source work, products, divisions, and ventures built for real environments.
Focused research becomes working software, then earns its path into a product, division, or venture.
Active division
Medical research infrastructure
A Kaelux division focused on secure AI infrastructure, research tooling, DevSecOps, and MLOps for retina, hearing, tinnitus, and medical imaging teams.
Research labs, hospitals, universities, foundations, and medical AI collaborators.
Public build
Unified AI studio for 3D professionals
A unified studio for 3D professionals that brings agent-assisted creation, Blender tooling, and production workflows into one workspace instead of another stack of disconnected subscriptions.
Built from extensive spatial-reasoning research into where AI struggles to understand, inspect, and reliably change 3D scenes.
Public build
Evidence-gated agent harness evolution
An open-source framework that helps agents improve task-specific harnesses through real attempts, artifact inspection, trace-backed diagnosis, candidate testing, and evidence-gated promotion.
First used to develop and measure the ViperMesh Blender harness; designed for coding, browser, visual, research, document, and custom application agents.
Public build
Prompt analysis and refinement
A prompt analyzer, refiner, and generator built around frontier system-prompt patterns, RAG-backed prompt references, and structured improvement workflows.
Builders, prompt engineers, and teams that need a repeatable path from rough idea to model-ready prompt.
Early-stage lab
Infrastructure security tooling
A local-first purple-team CLI for Terraform IaC with LocalStack sandboxes, red-blue reasoning, deterministic remediation, and evidence reports.
Infrastructure teams, DevSecOps builders, and security-minded platform engineers.
A fast terminal-style guide for venture questions, investor fit, partner builds, and business automations.
For businesses that see the Kaelux ventures and want comparable automation built into their own operations. We scope focused systems for intake, research, reporting, handoffs and other repeatable work around your real tools, data, and team.
Explore automationsWorkflow-first
Built around repeatable business workSystem-aware
Connects tools, data, and handoffsScoped
Automation without a bloated agency package
Kaelux is an Estonia-based AI and ML research lab and venture group. We study how intelligent systems reason, use tools, inspect artifacts, and operate in real environments, then publish the useful work or develop it into products, divisions, and ventures.
Founder and AI/ML systems researcher based in Estonia, developing Kaelux across medical AI, spatial reasoning, agent harnesses, prompt systems, and infrastructure security.
Kristofer's role is to turn research into working artifacts, open-source tools, and products, then build partnerships where a venture or secure business system needs hands-on engineering.
For investors, partners, collaborators, or businesses inspired by a Kaelux venture, send the context. Founder-led conversations start here.
Reach out to discuss investment, a venture partnership, MedAI collaboration, business automation, or a serious build.
