Copenhagen, Capital Region of Denmark, Denmark
Designing, building, and scaling the production infrastructure behind AI training, evaluation, and deployment pipelines used by millions. Working primarily in Python and TypeScript across backend services, APIs, and data pipelines, architecting for reliability and long term maintainability at scale on cloud and containerized environments. Collaborating asynchronously with ML engineers, researchers, and product managers across a global distributed team. Day to day this covers system design, code reviews, debugging across distributed systems, and contributing to engineering best practices.
Building a cloud native digital bookkeeping platform for professional service firms and creatives. I work primarily on backend development in .NET, with cross functional frontend collaboration, designing and building the microservices behind the product + service orchestration with .NET Aspire. Features I have delivered include Stripe subscription and seat billing, currency exchange rate feeds from Danmarks Nationalbank, CVR company lookup via Virk.dk, VAT reporting to the Danish Tax Agency through its NemVirksomhed (RSU) interface with automatic VAT return calculation from the standard chart of accounts, SAF-T import and export ahead of the 2027 audit file mandate. I also help shape the product's compliance direction, reviewing the Bookkeeping Act (Bogføringsloven) and its executive orders, translating the Danish Business Authority's digital standard bookkeeping requirements into implementation specifications, mapping out the IT security and risk assessment work required for registration, and researching e-invoicing over NemHandel (OIOUBL and Peppol BIS). Day to day this also covers project planning and bug fixing.
Oracle Status Worked on the projects "Clutch Zayu", "Maraca Camera", "Mechanic Astrologer", and "Meter Pavilion" (OpenClaw). Contributing to the development of advanced AI language models through specialized data annotation, prompt engineering, trajectory refinement, and quality assurance work at Outlier. Creating comprehensive system prompts that define AI model behavior, personality, and operational constraints. Designing and annotating single turn and multi turn conversational and agentic datasets that test model capabilities across diverse scenarios, skill categories, and risk domains. Evaluating and correcting model responses by identifying errors in tool usage, response accuracy, trajectory coherence, and adherence to safety taxonomies. Working with complex guidelines covering 50+ risk categories including algorithmic bias, misinformation, privacy protection, and ethical AI behavior. Serving as both an author and reviewer on Mechanic Astrologer, auditing peer submitted Task Blueprints against a structured rubric, verifying ground truth data, fixing issues end to end, and writing teaching feedback. Ensuring all training data maintains strict impartiality while testing models against real world use cases. Key responsibilities include crafting detailed system prompts with dynamic behavior scaling, designing autonomous agent trajectories that exercise skill discovery, memory usage, long horizon context, native tools (such as cron and sub agent delegation), and skill gap extension, annotating tool calls and text responses, refining Golden Trajectories for efficiency and grounding, and maintaining high quality training data standards. Work directly impacts AI model performance in areas such as factual accuracy, tool selection, safety compliance, agentic orchestration, and user experience optimization.
Developed and deployed REST APIs including authentication systems, rate limiting, and automated monitoring for external integrations, handling production level security and database management Built fullstack web applications using Angular, .NET, and Bootstrap, implementing features such as user authentication with JWT tokens, real time data visualization dashboards and automated reporting systems. Optimized and maintained enterprise resource planning (ERP) systems, including implementing automated email notifications, fixing critical bugs in production environments, and developing new features for time tracking and project management workflows