Overview
LLMday London 2026 is a one-day, in-person community conference for engineers building with large language models, AI agents and machine learning. It is part of a worldwide LLMday series, and it brings together practitioners who work on AI-powered products and systems. The goal is to share real-world experience, learn from peers and examine how software engineering is changing in the age of AI. The event expects 30+ speakers, three parallel tracks and 150+ attendees.
Date, time and venue
The conference takes place on November 26, 2026, from 9:00 AM to 3:00 PM at Everyman Canary Wharf, Crossrail Place, Canary Wharf, E14 5AR, London, UK. The venue is on Level -2. It is reachable by Tube on the Jubilee and Elizabeth lines, or by DLR, with Canary Wharf station as the nearest stop. Registration and coffee start at 9:00 in the main lobby, lunch and networking run at 12:00, and the day closes with a wrap-up at 15:00.
Format
The event is held in person and runs three parallel tracks, each on its own screen. Talks are scheduled at half-hour intervals, and the programme is marked as a work in progress, so times and slots may still change. Attendees can scan each other's QR codes at the close of each track and head to a nearby pub.
Agenda
The programme covers a broad range of topics across the three tracks.
Track 1 focuses on retrieval, healthcare, trust and evaluation. Sessions include a tour of interpretability in embedding models for search, designing LLMs for healthcare that clinicians can trust, production-grade prompt systems for agents, whether LLMs and agents can be trusted, how scientific ground truth can disagree, debugging code produced by agents, using satellite data and agents for parametric flood insurance, evaluating subjective LLM outputs, and putting AI to work inside enterprise systems of record.
Track 2 focuses on agent evaluation, production frameworks and applied engineering. Sessions include a layered approach to agent evaluation, the 1.0 release of LLM4S for Scala and the JVM, testing customer-service agents before they act, Graph RAG and small language models for regulated industries, migrating a virtual assistant to an LLM-driven orchestration platform, how agents interpret human requests, lessons from agents that caused real operational problems, agentic workflows in iOS development, and reproducibility in AI-driven scientific discovery.
Track 3 focuses on engineering practice, post-training and prompt management. Sessions include the relationship between engineering performance and AI systems, post-training a search agent in a synthetic environment, tools that shape what we can build, recursive prompt engineering in multi-agent systems, building a knowledge graph for personalised language learning, prompt release engineering, prompting within a creative production pipeline, lessons from post-training very large models, and a case for simpler, more deterministic enterprise AI.
Speakers and companies
Speakers come from companies and institutions including Optimizely, Caringdots, SoftNet Technologies, T-Labs, Red Buffer, LLM4S Foundation, The Trade Desk, AWS, Comply2Reg, IBM, Blindsight, FireHorse, inDrive, all.health, OTTRA, GitLab, Whitney AI, Université Paris Dauphine - PSL, Superagentic AI, Busuu, Wise, Beat Saber, Cursive and Great Wave AI, among others.
Who should attend
The conference is aimed at software engineers, machine learning practitioners, AI engineers and technical leaders who are building or evaluating LLM-based products, agents and AI systems. It suits teams that want practical lessons on evaluation, reliability, governance and production deployment.
Tickets and registration
Attendees can register on Luma. The organiser also invites companies to become sponsors, hosts or partners.
