Accelerating Collaborative Machine Learning

Abstract: 

The pace of recent innovation in AI is astounding, and adapting the research findings to new domains becomes a bottleneck. Whether repurposing, finetuning, or operationalizing published AI models in a different context, clarity on previous work is paramount. The more accessible and collaborative the process of developing machine learning applications becomes, the more efficiently we can make progress—leveraging the artifacts, incremental improvements, and insights of other contributors. However, as the global machine learning community grows and teams become more distributed, this crucial open collaboration becomes challenging. Common difficulties include onboarding new contributors, debugging across inconsistent setups, planning future development, and delivering reliable models for production or reproducible research. How can we standardize our datasets and model variants, track complex workflows, share insights, document the current state, coordinate next steps, and simply bounce ideas around without using ten different tools and occasionally losing progress? I will present tools and strategies—like systematic logging, dynamic queries for analysis, and flexible visualizations, all available in individual and group contexts—via case studies from computer vision for environmental sustainability, parsing structured documents to track money in politics, and deep reinforcement learning for generalizing safe agent behavior, to help inspire and streamline your future machine learning work.

Bio: 

Stacey Svetlichnaya is a deep learning engineer at Weights & Biases, building developer tools for accessibility, transparency, and collaboration in machine learning. Her research in computer vision and natural language processing includes image aesthetic quality and style classification, object recognition, photo caption generation, and language modeling for emoji. She has worked extensively on image search, data pipelines, productionizing machine learning systems, and automating content discovery and recommendation on Flickr, the first and longest-active photo-sharing website. Prior to Flickr, she developed a visual similarity search engine with LookFlow, a startup of 5 engineers which Yahoo acquired in 2013. Stacey holds a BS ‘11 and MS ’12 in Symbolic Systems from Stanford University.

Open Data Science

 

 

 

Open Data Science
One Broadway
Cambridge, MA 02142
info@odsc.com

Privacy Settings
We use cookies to enhance your experience while using our website. If you are using our Services via a browser you can restrict, block or remove cookies through your web browser settings. We also use content and scripts from third parties that may use tracking technologies. You can selectively provide your consent below to allow such third party embeds. For complete information about the cookies we use, data we collect and how we process them, please check our Privacy Policy
Youtube
Consent to display content from - Youtube
Vimeo
Consent to display content from - Vimeo
Google Maps
Consent to display content from - Google