AI Tinkerers Liverpool Meetup at FACT Lab

Overview

An in-person, hands-on meetup focused on Human-in-the-Loop (HITL) in ML/AI systems. We’ll host short talks and live demonstrations, with a strict emphasis on executable code, practical setups, and real-world lessons learned. Presenters will show what worked, what didn’t, and how HITL works in the lifecycle of model development and deployment.

Key details

• When: esday, November 18, 2025, 6:30 PM - 8:30 PM
• Where: The FACT Lab, 88 Wood St, Liverpool L1 4DQ, UK, Liverpool
• Capacity: 20 attendees
• Format: demo-first sessions with live code where possible
• Registration: open now on this page; slots are limited
• Submission: Submit your talk proposal via the form

What to expect

• Short talks and live demonstrations
• A friendly, welcoming environment for demos and quick, technical deep-dives

Submit Your Talk Proposal: Submit Your Talk Proposal

How to participate

• Propose a HITL-focused talk or demo or ‘What I Learned’ fast-technical-deep-dive (5–15 minutes).
• Provide a concrete setup: code, rough system design, and expected outcomes.
• Include what you learned and what others can replicate.
• Proposals are accepted through the official form linked above.

Sponsorship

The Liverpool Chapter of AI Tinkerers is sponsored by Prolific, the human intelligence layer for AI building and development.

💬 Prolific is a technology company building the biggest pool of quality human data in the world. And the ultimate platform to access it.

Over 35,000 AI developers, researchers, and organizations use Prolific to gather data from real, engaged people—with a new study launched every 2 minutes.

Our platform makes it quick and easy to get data you can depend on, from a pool of 200k+ verified and diverse research participants. So you can reach your next breakthrough faster.

Speakers and guests panels

📽️ Speakers Panel

🧑‍💼 Special Guests Panel

Demos and showcase

Great demos share a core pattern: they provide a concrete, executable setup within a short time frame, demonstrate tangible outcomes or metrics, and clearly articulate what was learned—what worked, what didn’t, and how others can replicate or adapt the approach. The speaker form’s emphasis on 5 to 15 minutes of technical deep-dives, executable code, rough system design, and outcomes is well aligned with what audiences reward: a reproducible pipeline, actionable steps, and candid discussion of tradeoffs. Demos should include a rough design or architecture sketch, sample code or configurations, data or metrics to validate results, and a clear takeaway or minimal replicable workflow. To avoid lower impact talks, presenters should ensure the content isn’t only high-level narrative or novelty without replicable steps, and they should anticipate potential issues and demonstrate resilience during the demo. In short, the strongest demos teach a clear HITL-enabled workflow that can be implemented by others, with transparent results and practical guidance to replicate or adapt.

Some concrete examples of the highest rated demos include:

• Sean Harris — I know Kung Fu’ - Downloading Brazilian JiuJitsu and Human-in-the-Loop, Liverpool, which focused on creating a BJJ dataset from competition videos and applying HITL labeling to train models for scoring and refereeing, illustrating a concrete, niche HITL pipeline with clear potential replicability.
• Sean Harris — A Noob’s Account of Learning about Basic Agent Testing with Deep Eval, Liverpool, which presented beginner-friendly insights into agent testing with DeepEval, highlighting approachable steps and practical, beginner-friendly guidance that attendees felt inspired to pursue.
• Sashank Pisupati — Automating error analysis - what works and what doesn’t, Liverpool, which offered a high-technical-depth exploration of error analysis with metrics like IAA and recall, including methods such as LLM Judge, axial coding, and Hungarian matching, demonstrating a robust, replicable experimental workflow.
• Carlos Fernandez Musoles — Put your idle devices to work and become your own AI cloud, Liverpool, which showcased an asynchronous LLM queuing system with just-in-time deployment and batch processing, highlighting real-world system design and practical performance considerations.
• Mark Walters — Conversational Agentic AI for UI control, Liverpool, which demonstrated a practical UI-control workflow via conversational AI, emphasizing an applied use-case and an engaging demonstration.
• Sean Harris — A Quick Demo of Vibe Kanban, Liverpool, which illustrated a conceptual system for orchestrating AI coding agents with a Kanban-like approach, valued for its clarity and potential for orchestration, even if it was more high-level.
• Kerry Parker — Build better with Cursor, Liverpool, which provided a hands-on walkthrough of building a functional full-stack productivity app with Cursor, appreciated for its practical, real-world coding focus and lessons learned.

How to stay in the loop

• Submit a talk proposal via the form above.
• If you’re accepted, you’ll receive a QR ticket for entry.
• Attendee eligibility and screening focus on practitioners actively building with AI; profiles linking GitHub/LinkedIn are encouraged.

Notes

Date/time and venue are based on placeholder details from the event configuration and the venue’s status. Final scheduling, talk slots, and access specifics will be confirmed for accepted attendees.

📥 Ready to register

Registration is open now. Space is strictly limited to 20 attendees due to the size of the lab :)