January 2022 Summaries

The Comet ML Office Hours, hosted by The Artists of Data Science, featured a session on January 26, 2022, focusing on various aspects of data management, including understanding, validating, versioning, and engineering data. The session included insights from Jimmy Whitaker of Pachyderm, Dr. Abe Gong of Great Expectations, and Data Governance Analyst Matt Blasa of Brinks Home Security, who shared experiences related to data curation, the emerging role of "Data Curators," and the potential pitfalls of not properly versioning data, illustrated by anecdotes such as issues with NLP models. Participants were encouraged to explore further resources through Comet's platform and to join future Office Hours to engage with experts and enthusiasts in the field.

The session discussed in the Comet ML Office Hours series, held on January 26, 2022, focused on data-related topics such as understanding, validating, versioning, and engineering data, emphasizing the critical role of good data in model performance. Key speakers included Jimmy Whitaker, Dr. Abe Gong, and Matt Blasa, who shared insights on data governance and the concept of "Data Curators," a term that led to a discussion about diversifying the "Data Scientist" title into multiple job roles. A notable story highlighted the challenges of data versioning, particularly in Natural Language Processing models that inconsistently output dates as digits or words. The session underscored the importance of proper data management in machine learning experiments and invited participants to explore further resources and discussions available on Comet's platforms.

In the third session of Comet ML Office Hours, hosted by The Artists of Data Science, researcher and professor Dr. Angelica Lo Duca emphasized the importance of baselines and tracking in machine learning experiments, noting that a good baseline should be simple, fast, and repeatable. The session, held on January 19, 2022, also focused on the significance of a standardized workflow and the reproducibility of experiments, which attracted Dr. Lo Duca to Comet. The discussion, led by Harpreet and Angelica, highlighted the value of experimentation, knowledge sharing within the community, and the need to constantly evolve by trying new models and tools. Both speakers underscored the importance of taking breaks during the learning process and encouraged participants to engage with Comet's resources, including Dr. Lo Duca's articles and work by notable figures like Andrew Ng. The Office Hours sessions, held weekly, are free and open to anyone interested in deepening their understanding of data science practices.

In the third session of the Comet ML Office Hours series titled "Seven Simple Steps to Standardizing the Experiment," Dr. Angelica Lo Duca emphasized the importance of using baselines to evaluate machine learning experiments and highlighted the necessity of a standardized workflow and consistent data tracking for reproducibility. She discussed how experimentation, iteration, and the exploration of new tools and techniques are vital for success in modeling, while also stressing the value of community knowledge-sharing. The session encouraged participants to engage with resources such as articles by Dr. Lo Duca and other experts, and to join the ongoing virtual Office Hours for further learning and interaction.

The Comet ML Office Hours, hosted by The Artists of Data Science, recently held Session 2 of their series on "Standardizing the Experiment," focusing on the importance of defining the scope and success in machine learning projects. The session emphasized the need for clear communication among stakeholders and the significance of understanding project objectives to ensure consistency across teams. Harpreet highlighted crucial habits for successful ML projects, such as tracking metadata and systematically logging data. The session also underscored the importance of asking detailed questions to align project goals and protect resources. Additionally, resources like Andrew Ng's Machine Learning Coursera course were discussed, alongside Harpreet's article on defining scope and success. The series is ongoing and invites participants to join future sessions, which are held weekly and are free to attend.

The second session of the "Seven Simple Steps to Standardizing the Experiment" series, hosted by Comet ML's Office Hours and The Artists of Data Science, focused on the importance of defining scope and success in machine learning projects. The discussion emphasized the need for clear understanding among stakeholders and consistent objectives across teams. Harpreet, one of the speakers, stressed the importance of tracking metadata and systematically logging data to ensure projects are reproducible and standardized. He also highlighted the necessity of asking comprehensive questions to gather essential information at the project's outset, referencing Andrew Ng's work as a valuable resource. The sessions are free and open to the public, with recordings available on YouTube, encouraging further exploration of the topics discussed.

The Comet ML Office Hours, in collaboration with The Artists of Data Science, recently launched a reimagined series focusing on standardizing experiments in data science, with the first session featuring a discussion between host Harpreet Sahota and Anaconda, Inc.'s Christian Capdeville. Key themes included the critical role of communication and timing in engaging stakeholders, particularly during the proof of concept stage, as well as the challenge of aligning machine learning projects with business objectives and team KPIs. Christian emphasized the importance of mapping processes to overarching project goals and referenced Teresa Torres’ book on creating customer and business value. The sessions are part of a larger series aimed at fostering community engagement and are freely accessible online, with more resources available through Comet's various platforms.