AI / cybersecurity / systems engineering

I build systems, test assumptions, and follow the interesting problems.

I am Jan Mueller, a multi-interested systems geek working across AI, cybersecurity, programming, DevOps, infrastructure, and hardware-adjacent experiments.

AI systems Cybersecurity Programming DevOps Infrastructure Agentic workflows

See Projects Collaborate

~/catdev/profile

$ whoami

security engineer / programmer / infra person / AI tinkerer

$ current-work --summary

Hikari

agentic monitoring, MCP device adapters, dark-fiber failure analysis

AXIOM

self-hosted RAG, research agents, PDF/web evidence, papers

cyb3r

AI security workflow tooling paired with a reasoning model

// Open to serious collaboration where systems, security, and AI meet.

profile

Systems, security, AI,
and the glue between them.

My work usually starts where domains overlap: agents that need real tools, infrastructure that has to survive failure, security research that should become usable software, and hardware that needs to be understood before it can be automated.

Systems that behave

I care about infrastructure that is understandable under pressure: Kubernetes, Linux, storage, Active Directory, backups, observability, incident response, and the operational details that decide whether recovery is calm or chaotic.

Practical security engineering

My security work is hands-on: security automation, lab validation, identity and infrastructure risk, web security, and the constraints that keep testing controlled and useful.

AI as engineering material

I use agents and LLMs as parts of systems, not as magic. That means datasets, tool contracts, eval loops, model behavior, policy layers, and the boring glue that makes AI useful in real workflows.

recent projects

Things I am building, maintaining, and testing.

A selection of current and recent work across dark-fiber monitoring, research automation, security AI, product prototypes, and open-source infrastructure tooling.

AI agent for OTDR monitoring

Hikari

Working prototype AI agent for OTDR monitoring

An autonomous OTDR fault-analysis layer for dark-fiber operations. Hikari ingests ALM telemetry, SNMP traps, REST data, OTDR traces, and event streams, then uses deterministic normalization plus agentic analysis to localize faults and draft incident packages.

How Hikari works

MCP adapter exposing OTDR device operations, alarm queries, port management, trace downloads, and trace analysis
Event loop for ALM alarms and operator messages, with baseline comparison, confidence scoring, and root-cause hypotheses
Firmware-derived ALM simulators for local testing of measurements, alarm transitions, monitoring cycles, and fault injection

MCP OTDR SNMP ALM Fault analysis Simulators

Self-hosted AI research

AXIOM

Research platform Self-hosted AI research

A self-hosted RAG and research-agent platform for document-heavy work. AXIOM can ingest PDFs and web sources, run agentic research missions, keep citations visible, and turn gathered evidence into structured research drafts and papers.

How AXIOM works GitHub repo

PDF and web-source research workflows with document Q&A, mission planning, notes, drafts, and writing assistance
Hybrid retrieval with vector search, BM25/OpenSearch, entity extraction, knowledge graph expansion, and citation profiles
Self-hosted deployment work includes GPU worker isolation, local hardware support, OpenAI-compatible APIs, and Go backend experiments

RAG OpenSearch pgvector GPU workers Go Research

AI-assisted security workflows

cyb3r

Active build AI-assisted security workflows

An AI systems project built around an operator-oriented reasoning model and a Pi-based tool layer. The focus is controlled security research workflows: planning, evidence handling, tool-use discipline, reassessment, and clear operator oversight.

How cyb3r works HF model card cyb3r repo

Pi-based tool integration layer for controlled lab workflows and repeatable operator actions
QLoRA fine-tuning work for a Qwen 27B model focused on branch ranking, global reassessment, and tool-use discipline
Training pipeline includes curated security corpora, policy rows, evaluation samples, and Hugging Face release preparation

Pi QLoRA Qwen Security AI Agent workflows Model training

Agentic product delivery

Theia

B2B SaaS PoC Agentic product delivery

A dark-fiber monitoring SaaS proof of concept designed and implemented end-to-end with agentic engineering workflows: product design, frontend, backend, ingestors, infrastructure, operators, and delivery artifacts.

Svelte/TypeScript frontend with Go backend services and admin workflows
Single-device ingestor model for ALM devices with gRPC control streams, health checks, and Kubernetes-friendly deployment
Platform pieces include infra, protobuf contracts, an operator, and an MCP device integration layer

Svelte Go gRPC Kubernetes OTDR SaaS

Maintainer Open source container tooling

Container-Compose

An open-source Swift tool that brings Docker Compose-like workflows to Apple Container. I currently help maintain it while it evolves around service orchestration, env files, dependencies, volumes, networks, and Apple platform behavior.

Project repo

Compose file interpretation for Apple Container
Multi-service orchestration and service dependency handling
Open-source maintenance, issue triage, review, and implementation work

Swift Apple Container Compose Open Source

background

Fifteen-plus years around production systems and security.

My work has crossed software engineering, cloud and infrastructure operations, storage, Kubernetes, Active Directory, security engineering, and recovery planning. That mix keeps me pragmatic: I like systems that can be reasoned about when they fail.

I also maintain and contribute to open source, study continuously, and use agentic workflows heavily in my own engineering process. I also like the hands-on edge of computing: small devices, lab networks, soldering, firmware behavior, and simulators that make hardware-dependent systems testable.

GitHub LinkedIn

Signals

Security background

OSCP, OSCE, CPTS, CWES, OSWP, hands-on lab work, security automation, and practical validation workflows.

Production operations

Kubernetes, Linux, storage, Active Directory, cloud operations, backup strategy, observability, and incident response across real production environments.

AI engineering

Agent workflows, tool contracts, RAG systems, fine-tuning datasets, evaluation loops, and local/self-hosted inference setups.

Hardware tinkering

Comfortable around the physical layer too: Raspberry Pi builds, network devices, lab hardware, soldering, firmware-derived simulators, and debugging with tools in hand.

15+

years in IT

OSS

maintainer work

lab and hardware tinkering

collaboration

Open to collaborate where AI, security, and systems overlap.

I am interested in serious technical work: agent tooling, security automation, infrastructure, product prototypes, research systems, and operationally grounded AI.

Start a Conversation info@catdev.io