Hi, I am Raffael. I live in Basel and work as an AI architect.

By day I am Enterprise Solution Architect for AI Platforms and Agentic AI at Helvetia Baloise Group. I help define a Group-wide AI platform: a common foundation for the secure, governed use of foundation models, agents, and AI-assisted engineering across market units with different technology strategies. Before that, I spent the first half of 2026 there as a platform engineer on secure web delivery and edge infrastructure, where a colleague and I built an LLM-assisted workflow that turns noisy ModSecurity alerts into review items an operator can act on.

Outside of work, I build automation-first platforms and edge systems at the intersection of security and AI, and a lot more besides: programming-language tooling, local-first apps, simulations, and games. For years my ideas lived in a notebook because there was never enough time to try them. AI moved the bottleneck from writing the code to deciding what is worth building, so now I ship ideas small and real and see which ones stick. What I write here is my own view, not my employer’s.

raskell=rascal×Haskell×Raffael(tanuki mascot included)

Tech that serves humans. Boring by default, precise when required.

I like hard problems. The kind where latency, reliability, risk, and people all collide. Applied security is just one arena of that. It is systems under pressure with adversarial inputs and questionable signal-to-noise.

Most of my ideas start with reading: books, papers, and long articles, many of them far from my field. I keep a reading log of what I am reading, what is next, and what I have read, and I review a book when I have something specific to say about it.

There is another thread, older than most of my projects and further from machines. For more than ten years I ran Personality Café Basel, a meetup about personality typology: the Enneagram, MBTI, Jung’s types read through their cognitive functions, and, with less conviction on my part, astrology. I also founded and still lead The Great Awakening Basel, a meetup for ideas from the fringe: religion and spirituality, alternative history, and the questions mainstream debate tends to leave out. In both, what holds my interest is less any single answer than the question underneath: how much a model, of a person or of the world, can capture, and what it leaves out. It is not so different from the question I ask about the models I work with today.

What I do

  • AI and agent platforms Model gateways, orchestration, context, memory, tool use, governance, and observability.
  • Platform automation Workflows, guardrails, self-service, operability.
  • Edge systems Reverse proxies, policy, observability, safe change.
  • Applied security Reduce noise, keep humans effective, ship pragmatic controls.
  • AI-assisted engineering Agentic workflows, model-driven tooling, ops automation.
  • Languages and tooling Compilers, package managers, frameworks, and specs.
  • Local-first apps Software that keeps your data on your device.
  • Simulations and games Interactive worlds, some that mimic real ones and some made for play.
  • Open standards Sustainable solutions over vendor lock-in.

Projects

I run several early-stage ventures and flagship builds, alongside a steady stream of experiments. The main ones:

  • Stable

    ZentinelZentinel

    Zentinel Security-first reverse proxy built to guard the free web. Rust on Cloudflare's Pingora, explicit configuration over hidden defaults, and WAF, auth and other policy in isolated agents, so a crashing agent leaves the proxy running. Agents for WAF, bot management, identity, API inspection and scripting, SDKs from Rust to Haskell, a pure Rust ModSecurity engine and a fleet control plane.

    Website
  • Open beta

    Archipelag.ioArchipelag.io

    Archipelag.io Decentralized AI compute: AI jobs run on a network of community-operated GPU hosts. Jobs prefer nearby hosts and a preferred region, and every job records where it ran; strict regional enforcement is in development. In free open beta: hosts earn virtual credits while metering is validated, and real-money payouts are not enabled yet.

    Website
  • Public alpha

    CyaneaCyanea

    Cyanea Bioinformatics notebooks that actually re-run. Write, run, and share analyses that execute in your browser on WebAssembly, with nothing to install. Every notebook, dataset, and output is a versioned snapshot you can run again. A Rust analysis stack underneath, 21 file formats built in. Open source, with self-hosting and federation on the way.

    Website
  • Shipping

    Arcanist Haskell tooling for the age of AI-authored code. hx is a fast package and project manager written in Rust, and bhc is a compiler built to check everything it can. With a registry that records what has been verified, they lead toward a commons where nobody, human or model, has to verify the same thing twice.

    Website

All projects → · Several of them are open to investment.

Speaking

I also founded and organised Cloud Native Basel, a meetup for people building on Kubernetes and cloud native infrastructure in and around Basel.

Contact

Everything I write here is free to reuse under CC BY 4.0, and the code samples under MIT. Credit me and link the page.

    Open the full search →