What we’re about
Meet with students, scientists, engineers, and entrepreneurs who are interested all things Data Science related! This is an exciting time for coders of every type in Leeds, so join us!
We are sponsored by Jumping Rivers (https://jumpingrivers.com), a predictive analytics training and consultancy company.
Upcoming events
1
2 Speakers - LLM Coding tools and Explaining LLMs with gSMILE
Platform, New Station Street, Leeds, GBWe are delighted to welcome you to our meetup in November! We have two speakers tonight: Dan from datavaluepeople and Zeinab Dehghani from the Edge AI Hub
If you would like to volunteer as a speaker please reach out to us at lds@jumpingrivers.com.
Schedule
18:00--18:45h: Refreshments (Food served on first come first served basis)
18:45--18:50h: Welcome
18:50--19:25: Dan @ datavaluepeople
LLM Coding Tools - Share & Learn
LLMs have already substantially changed how many of us write code; Copilot, ChatGPT, Claude, Cursor, Gemini, SWE-agent, Q Developer, Codex—the list goes on.In this session, Dan will present his own explorations into a variety of these tools, from basic copy-pasting with ChatGPT, to agents that generate full features and open merge requests from a single prompt. He will share what he found useful, what was a pain, how much they cost, and how easy they are to get started with. Crucially, you’ll all be invited to input and share your own experiences and opinions throughout!
By the end of the session, you should have a solid sense of the tooling landscape—and a clearer idea of which tools might be worth your time, and which to skip. Come along and bring your workflows with you!19:25--20:00h: Dr Zeinab Dehghani @ Edge AI Hub / University of Hull
Explaining Large Language Models with gSMILELarge Language Models (LLMs) such as GPT, LLaMA, and Claude have achieved remarkable performance in text generation, yet their decision-making processes remain opaque. This lack of transparency poses challenges in domains where trust and accountability are critical. In this work, we introduce gSMILE (Generative SMILE), a generalized and LLM-oriented extension of the previously proposed SMILE framework, for interpreting how these models respond to specific parts of an input prompt.
gSMILE is model-agnostic and operates by introducing controlled perturbations to the input, observing the resulting output changes, and identifying which input words have the most significant impact on the output. This process enables the generation of intuitive heatmaps that visually highlight influential words in the prompt. We evaluate gSMILE across several leading large language models (LLMs) using metrics such as accuracy, consistency, stability, and fidelity, demonstrating its ability to provide precise, robust, and reliable explanations. By making LLMs more interpretable, gSMILE takes an important step toward building more transparent and trustworthy AI systems.Bio 1: Dan
Daniel Burkhardt Cerigo is a full-stack machine learning scientist and engineer. He has been building end-to-end ML systems, leading data sci teams, consulting, and educating on applied data science for almost 10 years. He founded datavaluepeople in 2020, a collective of freelance data scientists and developers, who've delivered projects like real-time decision-making algos for optimising pharmaceutical production facilities, and an open source non-commercial social media data collection and analysis platform for peacebuilders (https://howtobuildup.org/phoenix/). When not typing into computers he likes listening to music and snowboarding.
Bio 2: Zeinab Dehghani
Zeinab is a Research Assistant at the University of Hull, working with the Edge AI Hub on federated learning. Her Master’s thesis focuses on explainability for generative AI, with an emphasis on transparency and safety. Her research covers AI safety, interpretability, and multimodal systems, and she specialises in large language models (LLMs), instruction-based image editing, and retrieval-augmented generation (RAG) pipelines. She has contributed to projects integrating text, image, and speech modalities and published work on explainable AI, including methods for transparency in LLMs and image editing models. With a background in statistics and over a decade of IT infrastructure experience, She brings a strong foundation for designing secure and scalable AI solutions. Her research is supported by the Centre of Responsible AI and the Dependable Intelligent Systems Centre.
News and Announcements
Have a news item or announcement you'd like to make about upcoming data events or job opportunities in Leeds? Comment below or contact us directly (lds@jumpingrivers.com) and we'll do our best to circulate this information at the end of the session.
Please contact the organisers if you would like to volunteer as a speaker for future events.
39 attendees
Past events
33
