Everything You Need to Build Retrieval and Ranking Systems

Components for Recommendations, Search and Chat

Thanks for signing up!
We'll be in touch 🙌
Oops! Something went wrong while submitting the form.
  • Streaming or batch ingestion

    Stream data to build and update user and item understanding in real-time.

  • Real-time feature store

    Streamline model development by unifying, storing, and sharing features.

  • Multi-connector SQL interface

    Pull all your data into Shaped without having to build your own ETL pipelines.

  • Model library

    Build state-of-the-art user understanding with LLM rankers, two tower architectures, tree models and more.

  • Continuous training

    Ensure models stay relevant with up to hourly continuous training.

  • GPU training

    Harness the power of GPUs for faster model training and reduced turnaround times.

  • Real-time inference

    Suitable for real-time applications that require < 200ms request latencies.

  • Cost efficient

    The most cost efficient way to host a scalable ranking model.

  • Scalable

    Dynamically scale to meet demand for 1000s of requests per second.

  • Offline and online metrics

    Evaluate model performance with offline and online metrics, and A/B testing.

  • User and item drill-down

    Gain deep insights with detailed analyses on user and item interactions.

  • Embeddings visualization

    Visualize and interpret your embeddings to understand user and items better.

Built by ranking experts

Inspired by the companies we previously worked at, built better, for everyone.

  • 💪  Double your engagement

  • 🛒  Increase conversions by 20%

  • 📈  Improve retention by 30%

Trusted by startups and the worlds largest companies

View Case Studies

“It only took us a week of working with Shaped to deploy a model that instantly increased engagement”

Jay Shea
CTO, Alter

"Shaped was easy to integrate, and our users are more engaged than ever."

Jon Vincent
VP of Engineering, Supergreat

“Shaped's understanding of our data and ability to create a custom model has been impressive. ”

Matt Koh
Co-CEO, Brandazine

“Shaped is the best of both worlds — easy to get started while allowing control over our features and models.”

Mishaal Al Gergawi
CEO, Axis

Designed for developers

Turn key end-to-end embeddings infrastructure for ranking.

  • 🤗  Easy-to-use APIs

  • 👩‍💻 Powered by PyTorch

  • ⚡️ Instant deployment

Create Model
Rank Request

We pull so you don’t have to push

Shaped connects directly to where your data lives.

  • ⌨️  Declarative SQL APIs

  • 🙌  Handles transforms in real-time

  • ✅  No manual work for you

Conquer the cold-start

Adapt every user-session in real-time.

  • ⚡  Re-rank in seconds

  • 🔌  Real-time data ingestion

  • 🏞️ Uses your unstructured data

Why Shaped?

  • Deploy in days

    No maintenance
  • Use all your data

  • White-glove support

    Private Slack
  • Secure by design

    SOC 2

Join the community

  • Twitter Link
  • Linkedin Link
  • Slack Link
Thanks, we’ll keep you updated  🥳
Oops! Something went wrong while submitting the form.

Get up and running in just 1 sprint 🏃

For developers

Waitlist for public API keys
You're on the waitlist! We'll be in touch 🙌
Oops! Something went wrong while submitting the form.

For companies 🏢

Schedule a demo with your data️
Thanks for signing up! We'll be in touch 🙌
Oops! Something went wrong while submitting the form.


How long will it take to integrate?

The integration process is quick. You can connect your data to Shaped in minutes, train your first model in hours, and fully integrate into your app in days.

How is Shaped different to Algolia, AWS Personalize...?

Prior to discovering Shaped many of our customers were using Algolia or AWS Personalize. To learn why they switched you can browse a short summary below, read a case study describing the challenges they faced or simply know that we have in many cases doubled the performance of these products.  

1. Multi-modal unstructured data understanding
Shaped uses state-of-the-art AI technologies like transformers and Large Language Models (LLMs) to make complex data types, including text, images, and video, usable for ranking systems. Using all types of your data significantly improves ranking performance. Algolia and AWS Personalize don't support unstructured data understanding.

2. Built for data scientists and ML engineers
Shaped connects directly to all of your data warehouses and application stores. This means no extra work for you and instant deployment via our CLI. In contrast, Algolia and AWS Personalize require manual transformations and for you to push your data. This adds extra work for the initial setup and ongoing maintenance.

3. Real-time ranking systems
Shaped specializes in real-time ranking systems. This means you can easily use ranking in dozens of places throughout your company, products and apps. Algolia primarily focuses on search, while AWS Personalize offers a batch based recommendation system.

4. White-glove support
We’re as hands on as you’d like us to be with your data, offering advice and helping you build your models.

For more information see these blog posts comparing Shaped with Algolia and Shaped with AWS personalize.

Can I use Shaped for multiple ranking use-cases?

Certainly! Shaped is designed to be used for all of your ranking use-cases. Typically the companies we work with deploy dozens. Once your data is connected to Shaped creating additional ranking models is easy. See our docs for more information.

How much data do I need?

There is no minimum amount of data required. Collecting interactions, for example clicks, views and or impressions is the only requirement.

For more information see this blog post How much data do I need for a recommendation system?

Why shouldn’t I build this in-house?

The high cost of hiring multiple machine-learning engineers, the long time required to build and the on-going full-time maintenance required. We handle scalability and reliability without the worries. Shaped can take you from 0 to 1 in a few days at a fraction of the cost.

What’s your pricing?

Our pricing is flat-fee monthly determined by usage. Once we understand your approximate number of monthly active users, items counts, and particular implementation details, we’ll be able to provide a pricing estimate.

How do you handle my data?

Shaped was built from the ground-up with security as a top priority. We operate as a cloud software-as-a-service (SaaS) platform, and only retrieve from your connected datastores when necessary to build your ranking algorithms. After training, most of your data is discarded, other than specific non-identifiable encoded features that are used at inference. We only require read access to your datasets and customer data never has to be persisted within Shaped. Furthermore, Shaped works on encrypted data if you need an added layer of security.

Shaped uses best physical, virtual, network and operational security practices. We rely on role-based authentication for all data access and records audit logs for every action. We avoid data replication and have clear multi-tenant isolation policies to ensure sensitive and critical systems are separated. Shaped uses isolated VPCs for all production deployments. Data is encrypted at rest using AES-256 encryption or higher and all ingress and egress traffic is encrypted via TLS 1.2+.

For more information see here.