London Research Community Launches With DeepMind And Microsoft AI Participation
AI Research Foundry London is launching a new paper discussion community to bring together researchers from academia, major technology companies, and startups. The initiative’s inaugural session is scheduled for Thursday, 22 October 2026, at Pillar VC in King’s Cross, London.
Shubham Jain, Lead Machine Learning Scientist at Sense Street, founded the initiative to fill a gap in the local tech ecosystem. “London has great communities for builders,” Jain explained in a LinkedIn post. “I couldn’t find one where researchers go deep on a shared paper.”
Data Creation Takes Center Stage At Inaugural King’s Cross Session
The opening session focuses on data creation, curation, and scaling, examining how training data shapes AI system improvements. Attendance for the 6pm to 8:30pm event is free, though registration has already sold out. Four researchers from academic institutions and commercial AI labs form the speaker lineup for the evening.
Joshua Placidi, a PhD student in Imperial College London’s Brain and Behaviour Lab, will discuss sequence composition in electronic health record foundation model pretraining. Google DeepMind Senior Research Scientist Yash Bhalgat, who works on Gemini Omni and served as lead captain for Omni 1.5 pretraining, is also joining the event. Microsoft AI’s Maximilian Croci will discuss the development of MAI-Thinking-1, drawing on his work with pretraining data that includes web-sourced mathematics, science, and code. Pierre-Carl Langlais, Co-founder and CTO of French-German AI lab Pleias, completes the speaker roster by examining a fully synthetic, single-stage approach to training large language models.
Comparative Compute Efficiency Drives The First Discussion Theme
The community’s first discussion draws its framework from September research by Dwarkesh Patel and Jerry Han titled Pretraining progress is mostly coming from data. Their experiments compared improvements in publicly available model training methods and datasets between 2019 and 2025.
At a training compute budget of 10¹⁹ floating-point operations, the researchers found that data improvements delivered a 12-fold gain in compute efficiency. Model improvements over the same period yielded a 3.7-fold gain, making data-driven gains 3.24 times greater in that setting. The authors caution that their experiments occurred at a relatively small scale and may not capture the full contribution of frontier model improvements, nor did they investigate whether synthetic data can expand training sets without hurting performance.
Future Meetups And Support From The UK Dynamism Fund
Pillar VC is hosting the opening event, with sponsorship provided through a Sparks Grant from the UK Dynamism Fund. AI Research Foundry London plans to hold meetings approximately once a month or every six weeks. Each upcoming session will focus on a single research theme and a selection of papers, maintaining free admission with limited places and curated attendance.
Frequently Asked Questions About AI Research Foundry London
Who founded AI Research Foundry London and why?
Shubham Jain, Lead Machine Learning Scientist at Sense Street, founded the initiative because he could not find a local community in London where researchers could examine shared papers in depth.
Who is speaking at the inaugural 22 October 2026 session?
The speaker lineup features Joshua Placidi from Imperial College London, Yash Bhalgat from Google DeepMind, Maximilian Croci from Microsoft AI, and Pierre-Carl Langlais from Pleias.
How is the new research community funded and supported?
Pillar VC is hosting the inaugural event in King’s Cross, while the initiative receives sponsorship from the UK Dynamism Fund through a Sparks Grant.
Can people still register to attend the first session?
No, registration for the 22 October event is now closed after selling out, though future meetups are planned every month or six weeks.
Keep reading