←
Justin Boylan-Toomey117 posts
Justin Boylan-Toomey avatar

Justin Boylan-Toomey

@jit_compile

Lead Machine Learning Engineer for @wellcometrust and @ExploreWellcome. Interested in machine learning, AI art and data science for good. (he/him)

◷ Joined December 2021▧ 7 Photos and videos
Justin Boylan-Toomey avatar
Justin Boylan-Toomey@jit_compile·22 Jun 2025

Climate Policy Knowledge Graph🌍Harrison Pim and Fred O'Loughlin from Climate Policy Radar gave a very interesting talk on how they constructed a knowledge graph, using #machinelearning to link policy documents to topics.

Justin Boylan-Toomey avatar
Justin Boylan-Toomey@jit_compile·22 Jun 2025

Particularly relatable was his description of the delayed feedback of leadership tasks such as relationships and team building verses the more immediate gratification of #programming type tasks. Well worth a watch when available, even if only for the great book recommendations!

Justin Boylan-Toomey avatar
Justin Boylan-Toomey@jit_compile·22 Jun 2025

Debugging Leadership🪲A fantastic talk by Matthew Upson of MantisNLP, in which he shared lessons from his personal leadership journey in the form of Python error types. Covering challenges such as delegation, communication and self care.

Justin Boylan-Toomey avatar
Justin Boylan-Toomey@jit_compile·22 Jun 2025

Python Engineering Excellence Workshop🐍An interesting workshop run by Sam Joseph, where we each shared approaches and tools for levelling up our Python skills. A lot of people mentioned pair programming, reading open source code and using LLMS.

Justin Boylan-Toomey avatar
Justin Boylan-Toomey@jit_compile·22 Jun 2025

Key takeaways: - Modularity is useful, combining structured PDF processing with traditional NLP models for improved workflow and performance. - Models are relatively new and getting better all the time. - There is now a plugin for Prodigy for working with PDFs!

Justin Boylan-Toomey avatar
Justin Boylan-Toomey@jit_compile·22 Jun 2025

Conquering PDFs💪Great talk by Ines from Explosion on SpacyLayout a new library that wraps IBMs Docling, a library for structured processing of PDFs. Breaking PDFs into objects, including processing table data!

Justin Boylan-Toomey avatar
Justin Boylan-Toomey@jit_compile·22 Jun 2025

Highlights✨Parallel PyTorch Inference🏃Great talk exploring the impact of free-threaded Python (PEP-703 ) on parallel inference of PyTorch models.

Justin Boylan-Toomey avatar
Justin Boylan-Toomey@jit_compile·22 Jun 2025

PyData London a 🧵Great weekend at PyData London, promoting DataKind UK and the great work they do bringing together volunteer data talent to help third sector orgs deliver impact using data.

Justin Boylan-Toomey avatar
Justin Boylan-Toomey@jit_compile·27 May 2025

Looking forward to the @pydatalondon conference in a couple of weeks, in a great new venue! If you get a chance do stop by the @DataKindUK stand to say hi and discover how we bring together skilled volunteers to help charities make the most of their data! pydata.org/london2025

Justin Boylan-Toomey avatar
Justin Boylan-Toomey@jit_compile·2 Apr 2025

Excited to share the new Wellcome Collection Knowledge Graph, enriching concept data to help people discover the breadth of the collection online. Developed through the work of talented colleagues on our MLE and Wellcome Collection teams. stacks.wellcomecollection.or… #museum