Junior AI Engineer · LLM Applications · Agentic Workflows · RAG
I grew up in the 1990s watching Napster rewire the music industry, the same industry I'd spend more than a decade working in.
Training as a musician taught me discipline and the patience to get things right. Later, as an Audio Engineer, my focus sharpened on something more specific: quality lives in the details you almost miss, and routing complex signal chains gave me an early intuition for how interconnected systems behave under pressure.
As an Operations Manager, I managed three studio buildings and led their Site Managers, working with IT, Finance and HR to untangle the recurring failure points behind slow rollouts and inconsistent service. That work gave me a real understanding of how businesses function and where they break.
Surrounded by the tools reshaping my industry, I made a deliberate decision to stop being the person who uses them and become the person who builds them.
I stepped away and reskilled in Software Development, AI Engineering and AWS Cloud Engineering, and recently completed a 530-hour AI Engineer internship at Kudan GmbH, building a conversational agent for their PRISM Cloud spatial data platform.
I'm currently deepening my knowledge of JavaScript and learning to build with React to round out the front-end side of that pipeline.
I'm now looking for a role where I can work across the full product pipeline, from idea validation to deployment.
Based in Berlin. English (working language), German B1.2, improving towards B2 and decided to get a C1 certification.
A natural-language agent that lets users query a spatial data platform: 12 read-only tools, per-request row-level security, 704 unit and 75 integration tests, and a 46-case deterministic evaluation harness. Integrated into the platform's QA environment by Kudan's team. Proprietary code, so there is no public repository. Reference letter available on request.
A multi-agent AI journaling app with mood-matched music recommendations, a chat assistant with semantic memory, and a fail-closed safety layer that stores only safe entries. Deployed to AWS with Terraform and since torn down; the infrastructure code and architecture are in the AWS branch.
A RAG assistant for a Berlin recording studio, built from the owners' requirements: it answers visitor questions from the studio's documents and saves leads to Supabase. Retrieval scored 1.00 recall@5 and 0.98 MRR on 21 cases. The public demo runs on synthetic data.
The MovieWeb application uses six RESTful routes to manage the full CRUD lifecycle for multiple users. It integrates API data and wraps the content in a high-contrast, digital retro aesthetic.
A RESTful Flask API for managing blog posts with CRUD operations, search, sorting, and rate limiting.