Rebecca Taylor - Navigating Large Companies to Build ML
Jun 13, 2023
auto_awesome
Rebecca Taylor, an experienced ML engineer with a PhD in Bayesian Inference, discusses navigating large companies to build ML. Topics include challenges of agility, risk management in banks, engaging non-technical staff for innovation, and integrating machine learning in big corporations. Also, she shares her climbing experiences in South Africa and Munich.
Navigating ML in large companies requires stakeholder management and buy-in.
Bayesian methods are chosen due to limited data and encoding domain knowledge.
Challenges in corporate data science involve system integration and compliance.
Balancing data science and platform engineering roles ensures project success.
Effective collaboration between data science and software engineering is crucial for ML deployment.
Deep dives
Rebecca's Background and Career Journey
Rebecca Taylor, a data scientist from South Africa, discusses her background in electronic engineering and computer science, leading her to her current position as the lead MLOps engineer at a large retail company. She highlights her transition from a large corporate to consulting and back to a big company, emphasizing her roles in data science and personalization.
Bayesian Statistics and Interest in Uncertainty Modeling
Rebecca delves into her fascination with Bayesian statistics, influenced by her academic exposure to probability theory and causal modeling during her undergrad days. She explains her choice of Bayesian methods due to limited data availability and the advantages of encoding domain knowledge into models, reflecting on the ongoing preference for frequentist approaches.
Transition from Data Science to Software Engineering Focus
Rebecca shares her journey from a data scientist role with a focus on production data science to a heavily software engineering-oriented position. She discusses the challenges faced in a corporate environment where data science projects often involve extensive work on integrating systems, APIs, and ensuring compliance with protocols.
Balancing Data Science and Platform Engineering
Rebecca expresses her daily struggle between choosing data science tasks and platform engineering responsibilities, ultimately gravitating towards the latter to ensure end-to-end system functionality and project success. She emphasizes the critical role of overseeing project architecture and collaborating with various team members to avoid potential project derailments.
Challenges of Collaboration in Large Companies for Machine Learning
Rebecca highlights the importance of effective collaboration between data science and software engineering teams in large organizations for successful machine learning deployment. She discusses the significance of clear communication, terminology alignment, and building mutual understanding to integrate machine learning models seamlessly into applications.
MLOps and Data Platform Maturity
Rebecca outlines the key components of mature MLOps, including experimentation tracking, model storage, data access, orchestration, package management, CI/CD integration, and monitoring. She emphasizes the significance of proper data versioning, seamless orchestration, and integration environments for effective ML engineering.
Indoor Bouldering and Outdoor Climbing Interests
Rebecca shares her passion for indoor bouldering and her history of outdoor climbing adventures in South Africa. While focusing on bouldering in Munich and introducing her child to the sport, she expresses a desire to explore outdoor bouldering in scenic locations like Innsbruck for a richer climbing experience.
Final Remarks and Engagement with Rebecca Taylor
The engaging conversation ends with mutual appreciation for Munich's outdoor activities, including mountain biking and climbing opportunities in the nearby Alps. Rebecca invites further engagement through her LinkedIn profile, encouraging those interested in MLOps and her career journey to connect with her online.
Podcast Transcript Summary
The podcast episode features Rebecca Taylor discussing her career journey from data science to MLOps engineering, her interest in Bayesian statistics, challenges in corporate environments, MLOps components, and personal insights on climbing and outdoor activities in Munich and beyond. Rebecca's emphasis on collaboration, data platform maturity, and balancing data science responsibilities provides valuable insights for listeners interested in the intersection of data science and engineering.
Rebecca Taylor has a ton of experience with ML, having worked as lead, and senior ML engineer, and holds a PhD in Bayesian Inference. We chat about making ML succeed in larger companies. This is a skill in itself, often requiring stakeholder management and getting buy-in. If you're in a similar situation, this episode is for you!
