About me
While an astrophysicist, my academic research focused primarily on methods for the analysis of telescope image data. I was particularly interested in reliable inference about weak gravitational lensing from large imaging survey experiments such as the Dark Energy Survey, NASA's Nancy Grace Roman Space Telescope, and ESA's Euclid. This meant that I used to think about noise, pixels, sampling, and galaxy shape inference a lot.
Since 2014 I have been working in the City of London for Fidelity, researching the potential use of new data sources, statistics and analysis techniques to improve investment decision making. Recent work has focused on highly scalable convex optimization strategies for investment portfolio construction, going beyond traditional mean variance approaches to increase stability and customization flexibility while minimizing turnover.
Prior to that, I led an interesting project that sought to use principles from behavioural economics and conviction narrative theory to design better tools and dashboards for Portfolio Managers. This work incorporated emerging NLP capabilities and built a proprietary investment research text corpus of over 10 million proprietary Fidelity and third party notes, reports and analyses, together with all regulatory and sustainability filings. In subsequent years this growing corpus has provided a variety of valuable NLP features for Fidelity investors, including sentiment analysis, topic modelling and thematic investment identification, and is now used in a range of Thematic funds including Systematic strategies which I manage with colleagues.
- My Linkedin employment profile and history can be found at https://uk.linkedin.com/in/barnabyrowe
- My ORCID page can be found here
- My Google Scholar profile page, and Google's estimate of my (very) slowly increasing h-index, can be found here
- An list of my accepted publications in peer-reviewed journals hosted on the Astrophysics Data System can be found here
- A selection of free preprints hosted on arXiv can be found here