Introduction

In an era of fragmented user attention, the need for highly personalized experiences has never been higher. Delivering relevant and customized experiences is paramount for increasing user engagement, conversion rates, and overall revenue. Shaped, an innovative AI (Artificial Intelligence)-powered platform, is designed to meet this need by providing a highly customizable and adaptable recommendation and search system. 

In this article, we will explore Shaped, including its applications in digital products and its integration with Flutter. We will also examine a case study to understand how exactly Shaped can help businesses.

What is Shaped?

Shaped is a powerful recommendation and search engine that leverages advanced AI models, including transformers and Large Language Models (LLMs), to process and rank complex, unstructured data like text, images, and video. The platform is designed for a wide range of industries, including e-commerce, social media, media platforms, and marketplaces, allowing businesses to optimize user interactions and drive meaningful outcomes.

Key features of Shaped include:

Easy Set-Up: Direct integration with existing data sources allows for rapid deployment without the need for extensive manual data transformation.

Real-Time Adaptability: Shaped can ingest and re-rank data in real time using behavioral signals, ensuring that the system remains responsive to user actions.

Model Library: Businesses can fine-tune LLMs and neural ranking models to achieve state-of-the-art performance tailored to their specific needs.

Explainable Results: In-session analytics and performance metrics help visualize, evaluate, and interpret the impact of the recommendations, ensuring transparency.

Secure Infrastructure: Shaped is built with enterprise-grade security, including GDPR and SOC2 compliance, to safeguard sensitive data.

Using Shaped In Digital Products

Shaped’s versatile platform can be integrated into various digital products to enhance user experience and drive business goals. Here is how it can be utilized across different sectors:

E-Commerce: By optimizing product recommendations based on user behavior, Shaped can increase conversion rates and average order values. The system can suggest products that are most likely to interest a user, thereby driving higher sales.

Social Media: Shaped can boost engagement by curating personalized content feeds that align with user interests. This turns passive traffic into an active community, enhancing retention and interaction.

Media Platforms: For media and content platforms, Shaped can transform user traffic into subscriptions by recommending articles, videos, or podcasts that match user preferences, increasing both engagement and revenue.

Marketplaces: Shaped can improve both the buyer and seller experience by optimizing search results and product recommendations, ensuring that users find what they are looking for quickly and efficiently.

Integration with Flutter

For developers working with Flutter, integrating Shaped into mobile applications is a straightforward process, thanks to the platform’s comprehensive API and SDK support. This is a basic overview of how Shaped can be integrated into a Flutter app:

1. Installing Shaped API

Start by installing the Shaped API client in your Flutter project. You can do this using the http package in Dart, which allows you to make HTTP requests to the Shaped API.

2. Configure the API Key

Once the client is installed, you will need to configure your API key, which you can obtain from your Shaped dashboard. This key will authenticate your requests to the Shaped service.

3. Making Requests

Use the Shaped API to send data about user interactions or queries from your Flutter app. For example, when a user views a product or clicks on a link, this data can be sent to Shaped to improve future recommendations.

4. Displaying Results

The ranked results returned by Shaped can be displayed in your Flutter app’s User Interface (UI). For instance, you might use the results to populate a list of recommended products, articles, or social media posts.

5. Real-Time Updates

Since Shaped supports real-time data processing, your Flutter app can dynamically update recommendations as the user interacts with the app, creating a seamless and responsive user experience.

Comparing Shaped to Competitors

1. Advanced Multi-Modal Data Understanding

Shaped utilizes state-of-the-art AI technologies, including transformers and LLMs, to handle complex data types such as text, images, and video. This capability allows Shaped to significantly enhance ranking performance by fully leveraging all available data. In contrast, Algolia and AWS Personalize do not support multi-modal unstructured data understanding, limiting their ability to process diverse data types and impacting their ranking effectiveness.

2. Seamless Integration for Data Scientists and ML Engineers

Shaped connects directly to data warehouses and application stores, enabling quick and straightforward deployment through its CLI (Command Line Interface)-based setup. This seamless integration minimizes additional work and complexity for data scientists and Machine Learning (ML) engineers. On the other hand, Algolia and AWS Personalize require manual data transformations and extra steps to push data into their systems, adding to the setup and maintenance burden.

3. Real-Time Ranking Systems

Shaped excels in real-time ranking, providing dynamic and adaptive recommendations that can be utilized across various applications and products within a company. This real-time capability ensures responsiveness to user behavior and enhances the relevance of recommendations, similar to what TikTok implements. Conversely, Algolia mainly focuses on search functionality, while AWS Personalize offers a batch-based recommendation system, which lacks the immediacy and adaptability of Shaped’s real-time approach.

4. Comprehensive Support

Shaped offers white-glove support, delivering personalized assistance and expert advice to help clients build and optimize their models. Shaped offers expert setup by a team of FAANG-experienced machine-learning engineers, ensuring that initial models are tailored to your business objectives, saving significant setup time and reducing the need for lengthy experimentation. The team prioritizes transparency, regularly sharing performance insights and ensuring that model results are interpretable and understandable. This high level of support ensures that customers can effectively utilize the platform’s advanced features. In comparison, Algolia and AWS Personalize provide less personalized support, which can be a disadvantage for organizations that need more hands-on guidance to fully leverage their recommendation systems.

5. Solving the Cold Start Problem

Shaped utilizes advanced machine-learning techniques and pre-trained models to address the cold start problem (an issue where the model cannot draw inferences for users which have insufficient data), allowing for effective operation with minimal initial data. The platform also supports unstructured data types, eliminating the need for manual metadata tagging, which enhances model accuracy and provides better signals for improved performance.

6. Pricing

Shaped differentiates itself from competitors like Algolia and AWS by offering a straightforward, flat-fee monthly pricing model. While Algolia charges $0.60 per 1,000 requests—leading to variable costs that can reach approximately $1,800 for 3 million requests per month—Shaped avoids the unpredictability of usage-based pricing. AWS, known for its complex pricing structure, requires detailed forecasting and ongoing monitoring to avoid unexpected expenses. In contrast, Shaped’s pricing is based on the number of users and forecasted usage, with two distinct tiers: a self-serve option, which is similarly priced to Algolia, and a premium tier with white-glove support. This approach provides customers with transparent and predictable costs, allowing them to focus on their core operations without the uncertainty of fluctuating monthly expenses.

6. Summary of Differences: Ease of Use and Maintenance

7. Summary of Differences: Technology

Case Study: Enhancing Diversity and Engagement on Breakr

1. About Breakr

Breakr is a platform that streamlines creator marketing within the music industry, connecting artists, creators, and record labels to promote new music. Founded in 2020 and headquartered in Atlanta, Georgia, Breakr has quickly gained traction, working with major industry players like Def Jam, Interscope Records, Meta, and Capital Records. With tools such as a smart matching engine, AI-driven campaign optimization, and automated relationship management, Breakr aims to help creators gain visibility and connect with audiences across all platforms.

2. The Challenge: Overcoming Popularity Bias

As Breakr grew, the platform faced a common challenge in recommendation systems: popularity bias. Popular creators were receiving disproportionate attention, while newer, lesser-known creators struggled to gain visibility. This imbalance was detrimental to Breakr’s mission of surfacing the next hot and upcoming artists. To foster a more inclusive environment and support the growth of new creators, Breakr needed to enhance the diversity of its recommendations without sacrificing relevance.

3. The Solution: Maximal Marginal Relevance (MMR) with Shaped

To address this challenge, Breakr turned to Shaped, which implemented the Maximal Marginal Relevance (MMR) algorithm to balance relevance and diversity in the recommendation system. MMR is a technique that optimizes the trade-off between showing highly relevant content and ensuring a diverse set of recommendations. By integrating MMR, Shaped helped Breakr:

Retrieve Candidate Items: Shaped retrieves potential recommendations using vectors built from user and item embeddings.

Filter Out Seen Items: Items that the user has already engaged with are filtered out to keep the recommendations fresh.

Score Candidates: Each item is scored based on a pointwise binary classification model that estimates the probability of user engagement or conversion.

Re-rank with Diversity Penalties: The final ranking of items is adjusted to penalize redundancy, promoting a more diverse set of recommendations.

4. The Impact: A 382% Increase in Diversity

The integration of Shaped’s MMR algorithm led to a staggering 382% increase in the diversity of creator profiles recommended on Breakr. This significant improvement had several positive outcomes:

Enhanced User Experience: Users began spending more time on the platform, engaging more deeply with content that aligned with their preferences, leading to higher retention rates.

Improved Match Rate: The increased diversity improved the match rate, helping users discover new and relevant creators that they might not have encountered otherwise.

Greater Engagement: By surfacing a broader range of content, Breakr was able to keep users engaged and interested, fostering a more dynamic and inclusive community.

From ideation to production, the entire integration process took just a few days, showcasing the efficiency and effectiveness of Shaped’s platform.

5. Next Steps: Personalized Search

Following the success of the diversity initiative, Breakr plans to further enhance artist and creator discovery by implementing personalized search features with Shaped. This will allow users to find and connect with creators even more efficiently, driving continued growth and engagement on the platform.

Conclusion

In conclusion, Shaped stands out as a powerful tool for enhancing digital products by offering personalized, real-time recommendations and search functionalities. Its ease of integration, including with frameworks like Flutter, makes it an ideal choice for businesses aiming to optimize user engagement and drive growth. Whether in e-commerce, social media, or any other digital platform, Shaped’s AI-driven approach can significantly elevate the user experience.