Autonomous Agents · Agent Federations · AI-Native Engineering
Building the next generation of AI-native systems.
I build autonomous agent systems that do real engineering work: an open-core agent platform written in Rust, federations of agents that take issues to merged pull requests, an AI engineering factory for a legal-tech company, and security research agents competing in Web3 audit contests.
About
I design and build systems where AI agents do real engineering work — not demos. My current work spans an open-core agent platform, autonomous agent federations that plan, implement, review and merge production code, and AI-driven security research on Web3 audit platforms.
Agent Platforms: Building Agentis — an open-core platform for autonomous agents: a Rust runtime, an agent DSL, and public colonies under Apache 2.0.
AI-Driven SDLC: Multi-agent "factories" that automate the software delivery lifecycle end to end, from issue intake to reviewed, merged pull requests.
Security Research: Autonomous bug-hunting agents active in audit contests on Immunefi and Sherlock.
Before going all-in on agents I spent 15+ years in deployment automation, DevOps and platform engineering — that operational discipline is what keeps my agent systems production-grade.
Experience
- Building an AI engineering "factory" — a GitLab-integrated multi-agent system that writes, reviews and ships production code.
- Designing agent workflows around merge requests, CI feedback and code review, with guardrails and QA gates before anything lands.
- Integrating LLM-driven code generation and review into an established codebase and delivery process.
- Ensuring high standards of software quality and security within the company's AI ecosystem.
Agentis · Founder & Lead Architect 03/2026 – Present
- Building an open-core platform for autonomous agents: a Rust runtime (~200k LOC), an agent DSL, and an evolution engine with capability-based security and cognitive budgets.
- Publishing ready-to-run federations in agentis-colonies (Apache 2.0) — colonies of daemon agents that learn from observation and gradually take over routine work.
- A development federation autonomously takes GitHub issues to merged pull requests — planning, implementing, reviewing and merging behind QA gates.
- Built sdlc-bench to measure the cheapest (model, effort) tier per engineering task class and calibrate agent routing with data instead of intuition.
flat-cyborg · Author 05/2026 – Present
- A PTY wrapper for driving LLM CLIs (Claude Code and others) from scripts and daemons — terminal screen-state extraction and structured output parsing, written in Rust with minimal dependencies.
- The code-generation backbone of the agent federations: every autonomous edit runs through it.
Web3 Security Research · dark-factory 06/2026 – Present
- Building an autonomous audit federation on top of Agentis: stateful invariant fuzzing, symbolic execution (Halmos), and fork-state verification against live deployments.
- Competing in audit contests and bounty programs on Sherlock and Immunefi; the pipeline has reproduced past High-severity contest findings end to end.
Ralph Platform · Lead Engineer 01/2026 – 03/2026
- Built a production-ready digital labor automation platform from scratch using AI-augmented engineering patterns.
- Implemented autonomous Git workflows with multi-agent coordination, production-grade guardrails, and cost management.
- Tech: TypeScript, GCP, GKE, Terraform, LangFuse.
FlowLint · Project Founder 11/2025 – 03/2026
- Built FlowLint, an open-source static analysis suite for the n8n ecosystem.
- Designed an AST-based linting engine (14+ security and reliability rules) for low-code environments.
- Shipped a full suite: CLI, GitHub App (BullMQ/Redis), Chrome Extension (React/Vite).
- Set up end-to-end CI/CD pipelines, automated publishing to NPM/Chrome Store.
- Designed and operated a self-service deployment platform with unified pipelines.
- Introduced Kubernetes and Istio, standardized deployment and traffic management.
- Implemented n8n workflows for onboarding/offboarding and GitHub Copilot usage reporting — inputs for budget/license planning.
- AI/LLM Adoption: Used Claude Sonnet for pair-coding and documentation in Backstage, writing AWS CDK and release documentation.
- Member of the software architecture team, closely collaborating with development, operations, testers and architects.
- Led DevOps for Air Bank's international expansion (DEV/UAT environments).
- Operations and automation — VMs, Docker, Docker Swarm, Kubernetes, Istio.
- Setting up CI/CD, monitoring and incident management across teams.
- Configured and monitored integration tests in Jenkins and Cucumber.
- Led a team of application specialists, set up processes and SLA/OLA metrics.
- Took over and extended deployment automation to TRAIN and UAT environments.
- Deployment coordination — PREPROD verification and PROD releases with go/no-go decisions.
- Designed a deployment platform for the EKIS side product — self-service, unified pipelines.
- Incident management, service requests, and RFCs for development.
- Deployed server applications on Oracle WebLogic and PL/SQL procedures.
- Performed deployments to TRAIN, UAT and PREPROD with go/no-go decisions.
- Operations monitoring and administration of IBM servers and infrastructure for enterprise environments.
- Subject Matter Expert for scripting (ksh, bash, PowerShell).
- Progressed from 1st-line to 3rd-line technical support for the AVG antivirus product.
- Pre-sales communication, forum support, and remote assistance via LogMeIn.
Projects
Skills
AI & Agents
Development
Security Research
Infrastructure
Languages
Education
Contact
The fastest way to reach me is email or LinkedIn. Based in the Prague metropolitan area, working remotely.