I build AI systems that are meant to be used.
Two years building the backend and infrastructure behind an industrial-IoT AI platform at Perinet, and a set of my own projects going deep on RAG, agents and LLMOps. I care less about the latest model and more about whether a system measurably works — and can show that it does.
I'm an AI Engineer focused on building systems that work in the real world — reliable tools people use, not demos that impress once and break.
At Perinet (industrial IoT) I've spent two years on the backend and infrastructure side of their AI platform: Python and Go services wired into real-time MQTT sensor streams, FastAPI APIs, Docker and Kubernetes, CI/CD. I also ran the model-benchmarking that shaped the platform's design decisions.
The deeper RAG, agentic and LLMOps work I've driven through my own projects: hybrid retrieval with RAGAs and MLflow evaluation, multi-agent systems on LangGraph, and an observability dashboard that traces every model call's latency and cost.
I'm finishing my M.Sc. in Artificial Intelligence at BTU Cottbus, with close to three years of professional software experience overall, including earlier enterprise work at Cognizant. Open to AI / ML / LLM Engineer roles across Germany.
AI Engineer (Working Student)
- Built Python and Go backend services connecting LLM workflows to real-time MQTT sensor streams — versioned FastAPI REST APIs, containerized with Docker and Kubernetes, with automated GitHub Actions CI/CD for zero-touch deployments.
- Led the containerization workstream for the company's AI platform — a chatbot, real-time MQTT anomaly detection, and a sensor-data exploration tool — handling profiling, QA, and deployment integration with the engineering and ML teams.
- Owned the model-benchmarking workstream: evaluated retrieval speed, generation quality, and trade-offs across model variants to inform the platform's chatbot and corpus design.
Software Engineer Trainee
- Built and maintained enterprise banking applications (COBOL, JCL, DB2) in agile sprints; developed structured debugging and production-deployment practices.
GraphRAG Agent
Builds a knowledge graph from documents, then answers multi-hop questions by traversing the graph instead of flattening it into chunks — entity/relation extraction, k-hop subgraph retrieval, grounded answers with citations.
knowledge graph · k-hop retrieval · cited answersGraphRAG Studio
The full-stack app over GraphRAG Agent: upload documents, watch a typed knowledge graph build live, then chat over it with k-hop subgraph retrieval and cited answers. Next.js front end with an interactive force-graph, FastAPI back end.
live graph viz · k-hop retrieval · cited chatMulti-Agent Research Pipeline
A 4-agent system (Planner → Researcher → Writer → Critic) built on LangGraph state machines that turns a question into a sourced report, fully automated end to end.
LangGraph · strict role boundaries · CI/CDRAG Evaluation System
A hybrid-retrieval RAG pipeline (BM25 + dense + Reciprocal Rank Fusion) with an automated RAGAs/MLflow evaluation harness and regression alerts on retrieval quality.
0.94 hit@5 · 0.96 citation presenceLLMOps Observability Dashboard
A self-hosted dashboard that traces every model call's latency, token counts and per-model cost across GPT-4o, Claude, Gemini and DeepSeek — no external tracing service.
full-stack · multi-stage Docker · 12 testsMultilingual News NLP Pipeline
An end-to-end German news pipeline: Whisper ASR, cross-lingual NER, fine-tuned event classification, translation and summarization — engineered to run on a single 4 GB GPU.
+13% F1 · 8.4× faster inferenceLLM Fine-Tuning — JD Extractor
Fine-tuned Qwen2-0.5B with QLoRA (4-bit) to extract structured JSON from job descriptions, training ~0.44% of parameters — reliable structured output from messy text.
100% JSON validity · <4 min on 4 GB GPUResume Tailor
A CLI tool I use daily that reads a job description, tailors a résumé and cover letter, and runs its own ATS and regression checks before producing the PDF.
multi-stage LLM pipeline · self-checkingAI & Agents
LLMOps & Evaluation
Programming & Backend
Infrastructure & DevOps
M.Sc. Artificial Intelligence
Focus: Machine Learning, Computer Vision, Explainable ML. Thesis on content-aware Vision Transformer optimization for efficient inference on edge devices (PyTorch).
B.Sc. Computer Application
Foundations in software development, data structures, databases and systems.
Open to AI / ML / LLM Engineer roles across Germany — remote, hybrid or on-site, and happy to relocate. The fastest way to reach me is email.