Working with Shaped
The Shaped team is composed of world-class machine-learning, data, and infrastructure engineers who have built real-time machine-learning systems at companies like Meta, Apple, Google, and H&M. Our team includes published AI researchers and data scientists who ensure your recommendation systems leverage the latest AI research. We take pride in our research-to-production pipelines, designed to deliver measurable results.
When you work with Shaped, we become your dedicated recommendation, search, or personalization team. You'll be assigned a point-of-contact personalization expert who will guide you through every step—from model creation to production deployment and ongoing optimization. As part of our white-glove service, we can integrate with your Slack workspace, organize regular meetings, and collaborate to ensure you consistently achieve your metric uplift goals.
Onboarding with Shaped
Our onboarding process is designed to get you up and running quickly. Here’s what to expect:
1. Goal Alignment 🎯
Before starting an integration, we align on your primary use-case and uplift goals. While Shaped supports a variety of use-cases, we begin with the one that offers the most potential for personalization uplift. For example, this might be your main feed or homepage discovery experience.
2. Integration 🔌
Together, we’ll identify the user, item, and event data sources that provide the most valuable signals for personalization. This includes determining the Shaped connectors you'll need. If a connector doesn’t exist, we’ll build it for you. We’ll also recommend best practices for logging and storing data to improve model performance.
Pro Tip: Don’t forget to log impressions!
Impressions are events fired when an item is shown to a user but not clicked. While not mandatory, they provide a low-confidence negative signal that helps recommendation systems understand what users might not like.
3. Model Creation 🛠️
We’ll provision your API key and create your initial model. This involves connecting to your data sources and transforming the data into the required format. We backfill historical data from your sources, ensuring the model is updated with new data from the start. The entire setup and training process typically takes 1-2 days.
4. Offline Evaluation 📊
Once your bespoke Shaped model is set up, you can evaluate it offline—without integrating Shaped’s rank endpoints into production. Using Shaped’s CLI, you can query our inference endpoints to manually evaluate results. We also provide visualization tools to help you analyze performance. This phase typically takes 1 week.
5. A/B Testing 🧪
Next, we’ll collaborate with you to run an online A/B test to measure performance uplift on a subset of production users. We’ll recommend tools and evaluation criteria to ensure accurate monitoring. This phase typically takes 1-3 weeks.
6. Production Roll-Out 🚀
Once we’ve confirmed that Shaped meets your metric uplift goals, the final step is a full roll-out to your user base. Let us know when you’re ready, and we’ll be available to assist with any issues during the process.
7. Model Iteration 🔄
The journey doesn’t end here. Shaped can support multiple personalization use-cases. Once you’re satisfied with your first model, we can repeat steps 3-6 for additional use-cases. Each iteration becomes faster as we refine the process.
Why Work with Shaped?
- Expert Guidance: Work with a dedicated personalization expert every step of the way.
- Fast Setup: Get your first model live in as little as 1-2 days.
- Proven Results: Achieve measurable uplift with state-of-the-art AI models.
- Scalable Solutions: Expand to multiple use-cases with ease.
Ready to get started? Request a demo or start exploring with our free trial sandbox today!