2025/12/05 –, Main Stream 言語: English
We've all been there: staring at a frozen app, a stalled progress bar. Annoying, right? But in medical robotics, a software malfunction can be the difference between success and failure during a surgical study.
In lung cancer diagnoses, a 3D Convolutional Neural Network (CNN) that is trained to assist is just a promise until a robust MLOps pipeline makes it a reality.
A model in a hospital isn't just about accuracy; it's about unwavering reliability and minimizing downtime in a secure, on-premise environment. This talk is a deep dive into the code that ensures that trust.
This talk is for anyone who has ever seen a brilliant model fail in the real world. We’ll explore a system built with Python’s ecosystem, designed specifically for a critical medical application.
You’ll learn how to tackle the most difficult challenges of turning a promising model—like those used in lung cancer diagnoses—into a trustworthy, reliable reality, even in a disconnected environment.
Public datasets used for training are often anonymized for patient privacy, which can make it challenging to verify a model's performance across a diverse population. I'll discuss this critical limitation as we explore how to build more robust and inclusive solutions.
What You'll Learn:
* Predictive Maintenance: Build a system that detects a component failure before it happens, using telemetry to monitor the "heartbeat" of a surgical robot
* Data Drift Detection: Ensure your model's accuracy doesn't decay as imaging data changes
* Model Optimization: Get a resource-heavy 3D CNN to run fast and efficiently, thereby reducing latency and ensuring timely results
This talk isn't about theory; it's about systems development. Join me to learn what it takes to design, build, and maintain a life-saving system for critical applications.
Lilinoe Harbottle is an Indigenous (Kanaka ʻŌiwi) Data Scientist who bridges algorithms, robotics, and healthcare. She leads AI initiatives at a San Francisco startup, developing advanced models. At Auris Health (Johnson & Johnson), she enhanced medical robotic systems, improving bronchoscopy and urology procedures. A champion for open source and inclusive tech communities, she is a Sequoyah Fellow of the American Indian Science & Engineering Society (AISES).