Introduction
RealPage is a leading provider of software and data analytics for the multifamily real estate and property management industry, delivering AI-driven solutions that enhance operations, optimize resident experiences, and drive business value.
Recently RealPage partnered with Wallaroo.AI to help them achieve their strategic AI goals of ROI, and instating proactive risk mitigation through dramatically accelerating time-to-market for AI models in production and enabling faster innovation, optimized operations, and significant cost savings while in control of model security and governance.
Production AI Opportunities for RealPage
RealPage saw an opportunity to make a series of additional AI production innovations to further accelerate value delivery from AI projects and investments. These innovations presented themselves as:
Accelerating & Scaling AI Production
The primary optimization area for RealPage was bringing AI production time lifecycle from 1-3 months to 1-3 weeks to enable the AI and Data science teams to accelerate product delivery with self-service tooling, while reducing unnecessary technical dependencies and eliminating operational bottlenecks from low-level infrastructure manipulation as a result of packaging and deploying AI models for inference in production. Compounding this, the AI team’s model changes required dependencies on disconnected development and Ops environments, causing lengthy deployment times and operational inefficiencies.
AI Infrastructure Optimization
RealPage’s extensive investment in their AI infrastructure, while designed to drive efficiencies and scale for their customers, presented an opportunity for cost optimizations pertaining to AI inference, as a result of running AI models to several disparate settings.
Increased Model Visibility & Governance
The value of data is realized when enterprises have the right insights to inform their decision making. At RealPage, once the AI teams managed to deploy their models to production, they started exploring tools to proactively monitor the performance of production models in real time to ensure best-in-class AI-enabled product experiences for their customers.
Wallaroo’s Solution
Wallaroo and RealPage worked closely to prototype, deploy, and scale AI initiatives. Wallaroo offered a great breadth of technology and AI solution expertise to the RealPage team to fully integrate the Wallaroo platform with RealPage’s AI/ML operations, leveraging Wallaroo’s core capabilities for AI production deployment, while ensuring a proactive feedback loop and governance over AI outcomes.
Deployment & Management
The Wallaroo AI Operations Center accelerated AI production by implementing a streamlined production process with automated security checks and quality control gates. This ensured model robustness and performance, while user-based access control permissions guaranteed that only authorized personnel could access the models. As AI production scales, built in workspace segregation helps teams manage their growing number of models effectively, thereby reducing bandwidth overhead and fostering tight collaboration and feedback loops.
Additionally, version management and control are essential for scaling AI production. By tracking model versions, teams can ensure that the best-performing model is deployed, streamlining the process and maintaining a clear audit trail.
Implementing these robust model governance mechanisms, including audit trails and version control, promotes transparency and accountability. This enables organizations to demonstrate regulatory compliance and mitigate risks.
With enterprise-grade model governance, security and access control capabilities in place, the Wallaroo team moved on to solidifying production speed and scale along with post-production model performance and ensuring efficiency, performance, and flexible adaptability. To achieve this, organizations must implement key strategies that optimize resource utilization and maintain model accuracy over time.
Wallaroo’s autopackaging capabilities streamlined the integration process of getting any model framework production-ready, accelerating time to production. By autopackaging and uploading custom model inference pipelines, RealPage data scientists quickly and easily removed typical integration dependencies and process bottlenecks, simplifying the process of taking models from development to a production-ready state.
Production & Inference
Wallaroo’s high performance AI inference engine further optimized and scaled the existing RealPage infrastructure for AI inference microservices. The Wallaroo AI inference engine is designed to help AI teams achieve their goals easily and efficiently through built-in data and memory management efficiencies, intelligent autoscaling, and concurrent batching capabilities. These features maximize hardware utilization and performance in large scale scenarios while delivering low inference and high throughput performance.
The RealPage team’s goals for streamlined AI production involved more than just drastically shortening the time from model development to deployment. They also wanted to equip their data scientists with the insights to ensure AI models were performing reliably in production. This was all part of their plan to achieve full lifecycle production AI efficiency.
Wallaroo facilitated easy comprehension and action on model performance data for the data scientists through detailed, centralized model monitoring, and easy to consume model analytics and visualizations. The built-in automation and continuous model monitoring offer real-time alerts and insights for early detection of performance issues.
Additionally, data scientists can easily set baseline thresholds to track model drift and accuracy, with comprehensive logging that captures all relevant data points, ensuring thorough analysis and easy accessibility regardless of where the inference endpoint is deployed. This real-time monitoring keeps the data scientists informed, while customizable alerts enable quick, proactive action and resolution, ensuring optimal model performance at all times.
The data scientists can also quickly optimize underperforming models using A/B testing and Shadow Deploy methods, allowing seamless comparison between updated and production models and in-place deployment of better-performing models without any production downtime.
In addition to this, scalability is enhanced through workload orchestration where the RealPage data scientists can easily automate and manage recurring AI workloads by ingesting data from predefined sources, running inference, and efficiently depositing results—all with the flexibility to scale to their evolving business needs.
Benefits/Business Impact
As a result of this deep collaboration, Wallaroo enabled a fast path to ROI for RealPage through a streamlined deployment process that reduces time to model production from 3 months down to an average of 9 days, and 40% average reduction in infrastructure costs. This is achieved through Wallaroo’s self-service capabilities for AI deployments, such as model autopackaging, deployment, automated monitoring and lifecycle management. The streamlined process empowers RealPage data scientists to scale model deployments and maintain optimal performance. Additionally, Wallaroo’s built-in automated, real-time monitoring insights enable quick, proactive action, establishing a tighter feedback loop with the development and Ops environment teams leading to streamlined AI lifecycle process and risk mitigation control.
“The beauty of leveraging Wallaroo for the production of data science models is the ease of transitioning a local prototype model/code base into its production environment. I’ve worked at companies where the production framework involves months of refactoring your code into abstract classes. With Wallaroo, you have two classes – one of which is simply your code base. This greatly reduces the time to production and empowers data scientists to transition their code into production frameworks earlier in the development process so they can leverage all the monitoring and additional capabilities that Wallaroo offers. As a result, your model is better tested and understood well before it officially enters production for your company.” Lead Data Scientist – RealPage.
