Börkur Sigurbjörnsson is an information scientist and data engineer with experience working as an applied researcher and engineer within academia, large corporate research organizations, and as a key member and essential contributor of early-stage startups in Iceland, Netherlands, United Kingdom, Spain, Germany and France.
Training, fine-tuning, and evaluation of task-specific machine learning systems (ML) and large language models (LLMs).
Development and evaluation of ranking algorithms, including their deployment as search engines and retrieval-augmented generation (RAG) applications.
Data pipelines (ETL + APIs): Focused web crawls, data extraction, ingestion into knowledge bases or other data stores, and deployment of data services via APIs.
Prototypes and minimum viable products (MVPs) for extracting data-driven insights and rapid business model validation.
Web crawling and information extraction (using LLMs) to build an explorable knowledge base providing insights into the impact of urban interventions and policy decisions (Urbanixm and Urban Data Stories).
Intelligent document processing to facilitate AI powered post-trade processing for financial institutions (Banqora).
Discovery of diverse data sources related to commercial real estate, ingestion into an interoperable data-format and serving via APIs (Placemake.io).
Machine learned estimation of reading level needed to understand a piece of information (WizeNoze).
Using machine learning to identify and disambiguate brand references in media articles (brandcrumb).
Interactive data-stories on a range of topics for various clients:
Analyzing the pricing of energy performance certificates in residential real estate.
Identifying opportunities for symbiotic CO2 recycling.
Exploring carbon-insetting opportunities in the fashion sector.
Generating knowledge graphs from diverse sources to enable faceted exploration of image search results [WWW 2010 | pdf] (Yahoo! Research / Yahoo! Image Search).
Facilitating exploration of Flickr photos through semantically structured tag clouds [summary | technical report] (Yahoo! Research).
Providing focused access to sub-article-level information from Wikipedia through a specialized search engine [DIR 2006 | pdf] (University of Amsterdam)
Get in touch on LinkedIn or drop me an email for more information about prices and availability.