Skip to content
View RaulMermans's full-sized avatar
🎯
Focusing
🎯
Focusing

Block or report RaulMermans

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
RaulMermans/README.md
Raúl Mermans — AI Engineer. Intelligence, Engineered. An animated constellation of connected systems. Applied AI Engineer, AI Agents, Developer Infrastructure

Portfolio LinkedIn All repositories

About

I'm an Applied AI Engineer building intelligent systems, developer tools, and software products that solve real problems.

My background combines engineering, product thinking, and business operations. I'm interested in what happens when AI moves beyond a prototype: how systems integrate with existing infrastructure, how agents behave under failure, and how we verify that something actually works.

I work from understanding a problem and designing its architecture to implementation, integration, testing, and evaluation.

Building something is only half the challenge. Proving it works is the other half.


Selected work

Eyes Without a Face ↗
Build for systems your AI cannot see.

A privacy-aware hybrid AI system that lets cloud models propose integrations without accessing underlying private records. Combines on-device inference, controlled disclosure, cloud reasoning, and local verification.

Hybrid AI · Gemini · Firebase · Privacy Engineering · Release candidate
Reeprov ↗
Reproduce the failure. Prove the fix.

Local-first debugging infrastructure for AI agents: execution capture, replay, causal comparison, evaluations, and reproducible regression checks.

Agent Reliability · TypeScript + Python SDKs · Public beta
Moody Blues ↗
Test what agents do. Verify what actually happened.

A controlled fault-injection environment that tests agent behavior against committed state using an independent Truth Oracle.

Python · Docker · Fault injection · 28/28 documented synthetic failures reproduced · Developer alpha
OpsTwin ↗
Test operational decisions before making them.

Decision-support software comparing staffing and workflow changes through paired simulation and uncertainty-aware analysis.

Next.js · FastAPI · SimPy · Demo
Cultural Elasticity ↗
Research cultural relevance with evidence.

Agentic cultural intelligence combining Claude, Qloo data, bounded investigations, and evidence-backed strategic analysis.

AI Agents · Qloo · Commercial intelligence · Research prototype
BI Notebook Lab ↗
Understand how DAX evaluates.

An in-browser analytical learning environment with an expression engine, filter propagation, semantic models, and executable grading.

TypeScript · 1,024 reported passing tests · 82/83 hand-verified cases · Demo

Engineering in progress

  • Open VS Code Agent — Local-first coding-agent desktop environment built around open-weight models, governed tool execution, and reproducible evaluation. A documented 24-task read-only benchmark improved from 37.5% to 75%; real-world desktop acceptance remains in progress.
  • IRIS OS — Personal AI operating system exploring specialist-agent orchestration, memory, context, and governed execution. Public repository is an architecture edition.
  • HALO Control — Local AI infrastructure for routing, model management, workload governance, and resource observability. Public architecture edition; runtime acceptance is incomplete.
  • AAO / Applied AI Operations — Private reusable foundation for governed AI workflows, approvals, execution controls, and integrations.

Beyond AI infrastructure

DemandOS · Relay · DataBrief AI · SearchSignal · Website Audit Agent

These projects explore forecasting, analytics, automation, and AI-assisted decision-making in commercial and operational settings.


Open-source contributions

Project Contribution Result
Lemonade Reject invalid backend-specific configuration keys Merged
OpenLIT LangGraph memory connector Merged
OpenJudge Fix degenerate empty-string grader matches Merged

Technical stack

TypeScript, Python, React, Next.js, Node.js, FastAPI, PostgreSQL, SQLite, Docker, Git, GitHub Actions, Vercel

AI Agents · Agent Orchestration · MCP · Local inference / Ollama · Evaluation & Observability · Playwright · Pytest · Vitest · SimPy · Applied ML · DAX


Engineering, in motion

Daily-generated GitHub contribution history with engineering milestones and verified upstream contributions

Regenerated daily from GitHub contribution data. External pull-request milestones are verified historical merges.

Upstream contributions: Lemonade ↗ · OpenLIT ↗ · OpenJudge ↗


FROM MODELS TO OPERATIONAL SYSTEMS.

Pinned Loading

  1. Reeprov Reeprov Public

    Local-first debugging and reproducibility infrastructure for AI agents.

    TypeScript 1

  2. moody-blues moody-blues Public

    MOODY BLUES — the Reality Distortion Lab for AI agents (Vultr Agent Rush 2026)

    Python 1

  3. IRIS-OS IRIS-OS Public

    Public architecture edition of a personal AI operating system: evidence-backed attention, governed agents, memory and recoverable execution.

    TypeScript

  4. OpsTwin OpsTwin Public

    Decision-support lab for testing service-workflow changes before deployment through paired discrete-event simulation, uncertainty analysis, and operational evidence.

    Python 1

  5. cultural-elasticity cultural-elasticity Public

    Cultural Elasticity — agentic cultural-intelligence product for the Qloo Agentic Hackathon

    TypeScript 1

  6. demand-OS demand-OS Public

    Demand forecasting and inventory-risk system that turns raw commerce data into forecasts, stockout signals, and human-reviewed reorder recommendations.

    Python 1