08.09.2026 aktualisiert
100 % verfügbar
AI Engineer & Founding Engineer | Agentic Systems, RAG, Knowledge Graphs
Freiburg im Breisgau, Deutschland
Nur Remote
Master of Science in Computer Science with specialization in AISkills
AI AgentsFull Stack Entwicklung
Description
I'm an AI-focused software engineer with an MSc in Computer Science (AI specialization) from the University of Freiburg and hands-on experience building production AI systems end to end — from agentic architectures and RAG pipelines to deployment and cloud setup.
Most recently, as founding engineer at an AI startup, I built an enterprise-grade agentic platform and an autonomous voice assistant for customer service, working directly with business clients from pilot through production. Before that I co-founded an EXIST-funded startup, and earlier built deep-learning web applications at General Electric.
What I bring to a project:
- Agentic AI & LLM apps — voice agents, multi-agent workflows, tool-use systems, telephony integration (VoIP/SIP)
- RAG & knowledge graphs — retrieval pipelines, LLM-based extraction, PostgreSQL/Apache AGE
- Full-stack & deployment — Python, Django REST, React/Next.js, Docker, AWS, CI/CD
- Product mindset — I've built things people actually use, not just prototypes
Erfahrung
Berufserfahrung0 - 5 Jahre
Freelancer seit2026
Sprachen
DeutschgutEnglischMuttersprache
Projekthistorie
Architected an enterprise-grade agentic platform handling high-volume deployments with custom IT integrations for business clients. Built an autonomous 24/7 AI voice assistant for customer service using agentic frameworks and telephony protocols (SIP). Engineered data pipelines for intelligent insights and partnered with clients through pilot phases to fine-tune AI to their brand voice and deliver production-ready deployments.
Built a modular, config-driven pipeline that turns any text dataset into a queryable knowledge graph using a local LLM and PostgreSQL + Apache AGE (openCypher inside Postgres). Constrained entity and relationship extraction to a declared YAML ontology. Designed each stage (load, chunk, extract, fuse, store) behind abstract interfaces for pluggable LLMs and graph stores, with unit and integration tests. Technologies: PostgreSQL, Apache AGE, Docker.
Created a benchmark on Kolmogorov Arnold Networks and compared that to Multi-Layer Perceptrons for reinforcement learning on the gym environment. Tested on three different gym environments and made a KAN based LSTM to test long term dependencies. Technology: PyTorch.