Systems that tell you whether something is achievable — before it costs you anything.
I build applied AI systems that turn uncertain human workflows into structured, testable tools. My work sits between AI deployment, LLM evaluation, operations automation, and product prototyping. I am strongest where a team needs someone who can understand a messy real-world problem, model the logic, build a working system, and explain it clearly to technical and non-technical people.
Best fit: AI implementation, LLM evaluation, operations automation, technical product, or solutions engineering roles.
MILLE
In validationAn ML workflow assistant that turns a chosen task and dataset into a runnable prediction workflow, a report, and a working demo — with a deterministic gate that catches broken specs before anything runs. Best framed as ML automation and AI assistant prototyping, not a foundation model.
JNGA / JENGA
Prototype / startup projectAppointment optimization for clinics and service businesses. Reduces no-show damage using reminders, cancellation windows, waitlist movement, and cascade rescheduling to recover empty slots — with an explainable risk score behind every decision.
PHYSIS
Working prototypeA hotel operations feasibility engine that models rooms, floors, cleaners, travel time, inspections, and maintenance blocks. It tests whether a hotel's daily plan is physically achievable before staff are assigned, using deterministic scheduling, CSV intake, validation, and what-if replanning.
ALMIRAAH
Research releasedAlmiraah is an MCP server built on classical Arabic root-and-pattern morphology. It maps words and their derivations into hyperbolic space, so the geometric distance between two forms stands in for how close their meanings are. The aim is to give AI systems a measurable layer of meaning to work with, instead of only statistical pattern-matching over text.
I'm looking for roles where I can help teams deploy practical AI systems, evaluate language models, automate internal workflows, or translate operational problems into software. The best fit is a hybrid role across AI implementation, solutions engineering, LLM evaluation, technical product, or operations automation.
AI Implementation
PHYSIS, JNGA, MILLE — going from a messy operational problem to a working, deployed system.
LLM Evaluation
Almiraah — Arabic NLP depth, semantic analysis, and evaluation across language and culture.
Operations Automation
Hotel and healthcare workflow systems — scheduling, feasibility, and staffing logic.
Technical Product
Translating an idea into a usable demo, and explaining it to technical and non-technical people alike.
Across MILLE, JNGA, PHYSIS, and Almiraah, the same method repeats: take an uncertain, unstructured problem — a machine learning task, a booking calendar, a hotel floor plan, a body of classical Arabic text — and translate it into a structured system that can be tested rather than just trusted. That instinct has roots in classical Arabic philology, a discipline built around translating meaning precisely across forms: root to derivation, structure to sense. The engineering work applies the same instinct to operations and language models — turning a messy, human description of a problem into something that can be verified, and explaining the result back in plain terms.
Available for AI implementation, LLM evaluation, operations automation, and technical product roles in Vienna or remote Europe.