Börkur Sigurbjörnsson

Freelance applied scientist applying geospatial analysis and information-retrieval systems to climate and sustainability challenges for early-stage ventures, alongside contract work building production search/extraction systems. • Built a geospatial siting tool matching clean-electricity generation projects to fashion-brand suppliers, to explore carbon-insetting opportunities, allowing the founding team to test the viability of the business model and pivot. • Analyzed OpenStreetMap data to map supply and demand for industrial CO2 recycling, informing business-development direction for a hardware startup. • Fine-tuned and evaluated LLMs for key-information extraction against a curated gold dataset; built RAG-based search over the resulting corpus. Delivered a working system in 8 months, used to demo to investors ahead of a seed round.

Solo-founded and built an urban knowledge discovery engine — combining web crawling, LLM-based content curation, and retrieval-augmented search — giving urban professionals structured access to intervention-impact knowledge, alongside custom analysis and research for sustainable urban development. • Built the full pipeline: crawling (Nutch), LLM-based content classification (PyTorch, scikit-learn), chunking and dense vector indexing (LangChain, pgvector), and a RAG search/browsing interface. • Delivered custom analyses on request, e.g. energy performance certificate pricing and BREEAM-certified neighborhood impact, and published data-driven insight pieces on urban topics (walkability, bike-share adoption, speed limits and traffic safety). • Ran customer development (30+ interviews), with several iterations to shape the business offering.

Principal Data Engineer building urban-data intelligence for real-estate investment—turning 15+ heterogeneous sources into 250+ time series and 10+ combined indices, exposed via APIs and search. Co-developed the core insight system, coordinated agile engineering in a four-person team, enforced data/search quality through systematic evaluation.

Data Scientist and Team Lead (7-person DS/engineering team) delivering urban-data products for city decision-making. Set team vision and orchestrated project delivery. Served as technical lead for integrating heterogeneous urban datasets and building insight models across 3+ national and European projects.

Backend Engineer building ML-driven content classification for a children’s search engine. Developed and evaluated Dutch reading-level models, productionized them behind APIs, and outperformed commercial alternatives—removing vendor dependency for core technology.

Engineer & Scientist building brand-entity recognition for news at scale. Developed and evaluated language-agnostic brand-name detection and disambiguation models (leveraging Wikipedia/Wikinews), and deployed them via APIs to insert context-aware brand links—enabling rollout across English, Spanish, and Catalan without proprietary training-data pipelines.

Applied Scientist in industrial IR/data mining, bridging research and production to ship search experiences at scale. Built ranking and faceted exploration for points of interest in Yahoo! Image Search, reaching millions of users; also prototyped semantic browsing UX (interactive tag-clouds of related entities) as exploratory research.