Apartment List, an online apartment rental marketplace, has appointed Lawrence Kennedy as Head of Product and Greg Moore as Head of Design and Research, both starting in Q1 2026. The hires signal the company's strategic shift towards building an AI-powered platform that serves renters throughout their entire rental lifecycle, not just during the search phase. Kennedy joins from Warner Music Group, bringing experience from senior product roles at Spotify, Pandora and YouTube. Moore comes from Meta, where he worked on AI and virtual reality, with previous positions at Google and Shopify. The appointments reflect Apartment List's investment in personalisation and AI to improve the rental matching process. Founded in 2011, the platform currently lists over 7 million rental units and has helped millions of renters find homes.

How Winder.AI helped Apartment List Eliminate data Drift and scale MLOps automation. Apartment List, a leading online rental marketplace, wanted to accelerate its use of machine learning (ML) to power smarter recommendations and better lead quality. But model deployment was slow, data pipelines were inconsistent, and engineers were heavily involved in every release. To modernize its machine learning operations, Apartment List partnered with Winder.AI to deliver specialized MLOps architecture and automation. Winder.AI built a scalable, self-service ML platform that unified data, reduced deployment effort, and improved overall reliability. The challenge. Apartment List's ML workflow had challenges in two main areas: * Huge discrepancies between training and inference data, leading to unreliable model performance once deployed. * Long production cycles, often taking months to deploy a new model because of the complex handover between data science and operations teams. These issues slowed iteration, limited experimentation, and increased engineering overheads. Business context. Machine learning sits at the core of Apartment List's mission to help renters find their perfect home. As data volumes and model complexity grew, the company needed modernized operational foundations. "Enable data scientists to deploy and monitor models independently, while maintaining enterprise-grade governance and consistency." - Steve Kim, Senior Engineering Manager, Apartment List * Accelerate deployment and retraining through automated workflows * Eliminate data inconsistencies between development and production * Introduce continuous validation and monitoring * Enable self-service capabilities for data scientists * Deliver a roadmap to guide long-term MLOps maturity Phase 1 - discovery & assessment. * Workshops and interviews across data science and operations * Mapped ML pipeline and pinpointed data drift sources * Identified bottlenecks delaying deployment Phase 2 - architecture design & methodology. * Migrated to unified feature store via Chalk * Implemented Kubeflow Pipelines for automated training * Explored long-term Metaflow integration * Defined governance, validation, and monitoring standards Phase 3 - roadmap & enablement. * Delivered phased roadmap and ownership boundaries * Trained internal teams for independence * Provided playbooks ensuring reproducibility and scalability Technology & platform foundations. * Unified feature store ensuring identical training and production data * Automated Kubeflow pipelines for training, validation, and deployment * Version-controlled workflows * Integrated monitoring and validation for drift and performance Operational governance. * Single Source of Truth * Reproducibility * Continuous Validation * Team Empowerment * Incremental Maturity Challenges overcome. * Misaligned training vs inference datasets * Manual, engineer-dependent deployments * Lack of shared ownership between teams | Result Area | Before | After | Impact | | Data Consistency | Drift and mismatch | Unified feature store | No training/inference drift | | ML Pipeline Robustness | Manual, ad-hoc | Automated Kubeflow pipelines | Reproducible training | | Production Path | Slow, manual | Streamlined CI/CD | Faster, low-touch releases | | Team Empowerment | DS relied on engineers | Self-service ML workflows | Full DS ownership | ROI & Impact. * Reduced model deployment time and effort * Faster experimentation and iteration * Scalable ML delivery with fewer engineering demands * Improved reliability in production models Customer feedback. "Winder.AI guided us toward a unified architecture and an automated deployment process. The structured discovery turned abstract concerns into concrete solutions." Recommendation Score: 10 / 10 - Steve Kim, Senior Engineering Manager, Apartment List Next steps. * Expand real-time personalization via feature store * Automated retraining on performance triggers * Strengthen governance and lineage tracking Key takeaways. * Unified data eliminated drift * Pipelines automated training and deployment * Data scientists own full ML lifecycle * MLOps maturity accelerated innovation Why Winder.AI. "Winder.AI created a clear path to a single source of truth... their structured discovery bridged our DS and Engineering teams and enabled faster, self-service deployment." - Steve Kim