Can you use AI to vibe-code a computer vision pipeline? Yes... sort of.
In December 2025, a group of friends built an art installation for Lost Paradise music festival. We built the bulk of the project in two weeks, after work hours, during late night hackathons in someone's living room. There is no way we would have been able to do this without Claude Code. That being said, computer vision is an interesting case such that most of the debugging happens when you press play and run the system live - and AI tools can't sit and watch video streams in the same way a human can. It is difficult to describe with words the exact results you're looking for in the output, in a way that Claude can meaningfully understand and act on. As a result, we came up with clever ways to share our outputs and problems with Claude, and learned a lot about its capabilities and limits in image processing and computer vision tasks.
This talk will take you through the process - from running pose tracking models on AI cameras, using OpenCV for camera calibration, iterating on DepthAI pipelines, creating real-time data visualisations to track keypoints, and what we learned about the limits of generative AI models when debugging real-world systems.
This talk will cover the following topics
Charli Posner is an AI engineer exploring the limits of modern AI models in real-world systems.
At Stile Education, she builds production AI systems - from pipelines that scan handwritten student work into the platform to image editing workflows for illustrated characters. She also develops infrastructure for evaluating and improving LLM outputs across product features and internal tools.
Her work spans LLM pipelines, vector databases, computer vision, and deep learning, including research in human pose estimation at Toshiba and the University of Bristol. She focuses on the practical challenges of deploying AI systems: handling noisy inputs, unpredictable outputs, and making models reliable in real-world applications.
Outside of work, she builds creative AI projects such as interactive pose-tracking installations and writes technical blogs documenting experiments with emerging AI tools.
Will Cohen is a full-stack Software Engineer based in Melbourne/Australia. As a senior engineer at Stile Education, he leads initiatives focused on scaling pedagogical content management. Over the past 8 years, Will has successfully solved large-scale problems in e-commerce search, education, banking, content moderation, and defence.
Known for his unique combination of skills in working as both a Product Engineer and Applied Scientist, Will is passionate about turning frontier AI - Bayesian methods, LLMs, and agentic systems - into reliable products that people trust.
Outside the professional realm, you can find him climbing rocks, catching waves, or recovering from injuries.
