This guide is structured from three perspectives. Find the one that matters most to you:

• For the Machine Learning Engineer: A look at the trade-offs between a glass-box experimentation platform and a black-box API.
• For the Engineering Leader & Developer: An analysis of a modern ranking framework versus the "glue code" tax of a siloed system.
• For the Product Manager: A comparison of a search utility designed for features versus a growth engine designed for KPIs.

At a Glance: Two Philosophies of Discovery

For the Machine Learning Engineer: A Glass-Box Platform vs. a Black-Box API

For an MLE, the core trade-off is between the ease of a configured system and the power of a controllable one. Algolia provides a highly optimized but more opaque API, which is ideal for teams who want to set it and forget it. Shaped, in contrast, is a transparent, "glass-box" platform designed for MLEs who need to own, understand, and iterate on the core model logic (note: Shaped can still be set and forget).

MLE Perspective: At a Glance

A True MLOps Workflow: "Relevance as Code"

With Shaped, your entire ranking model is defined in a version-controllable file. This is "Relevance as Code." It fits naturally into your existing CI/CD pipelines.

1 model:
2   name: hompage_search_and_recommendations
3   description: "Recommend items to show in the for you feed"
4   pagination_store_ttl: 600
5   policy_config:
6     scoring_policy:
7       policy_type: xgboost
8     embedding_policy:
9       policy_type: item-content-similarity
10 
11 connectors:
12   - id: items
13     name: items
14     type: Dataset
15   - id: users
16     name: users
17     type: Dataset
18   - id: views
19     name: views
20     type: Dataset
21 
22 fetch:
23   events: |
24     SELECT user_id, item_id, label = 1
25     FROM views
26     WHERE event_type = 'view'
27   users: |
28     SELECT user_id, item_id FROM users
29   items: |
30     SELECT user_id, item_id FROM items

Offline Evaluation as a Prerequisite for A/B Testing

A/B testing is crucial, but it's expensive and slow. Before you even decide what to test online, you need to evaluate models offline. Shaped provides robust offline evaluation metrics for every trained model, including Precision@K, MAP, and NDCG. You can compare these metrics between model versions directly in the dashboard, allowing you to iterate quickly and make data-driven decisions about which models are worthy of a production A/B test.

Solving for Real-World Complexity

Our ranking framework is designed for advanced, real-world complexity. For example, you can create sophisticated multi-stage ranking models that combine an initial semantic search retrieval with a deep, personalized re-ranking step, all within a single, unified system. This allows you to build experiences that are both broadly relevant and deeply personal.

For the Engineering Leader & Developer: A Modern Ranking Framework vs. The 'Glue Code' Tax

For an engineering leader or developer, the decision comes down to system architecture, maintainability, and total cost of ownership (both in dollars and developer-hours). Algolia's architecture, with separate products for search and recommendations, requires your backend to be a complex orchestration layer. Shaped provides a unified framework that handles this complexity on the platform, simplifying your stack.

Engineering Perspective: At a Glance

Flexible and Simple Integration

We believe that connecting your data should be easy, regardless of your current infrastructure. Like Algolia, we offer a simple, easy-to-use Push API. For teams with a modern data stack, we also provide direct connectors to data warehouses and event streams like Snowflake, BigQuery, and Kafka, allowing for even richer feature engineering.

A Developer Experience That Respects Your Time

Algolia's InstantSearch libraries are excellent for building traditional search UIs. We acknowledge that. For teams that value control and a clean workflow, Shaped is API-first. We don't lock you into a UI paradigm. Our SDKs and APIs are designed with ergonomics in mind, making them a joy to use.

Here are practical examples of fetching results using our SDK and architecting for resilience, a pattern we guide all our clients on:

# Install SDK
pip install shaped

import shaped
api_key = 'your_api_key'
client = shaped.Client(api_key=api_key)

response = client.rank(
    model_name='hompage_search_and_recommendations',
    user_id='1',
    limit=5,
)

for item in response.metadata:
    print(f"- {item['title']} (Category: {item['category']})")

// Install SDK
npm install @shaped.ai/client

const shapedai = require('@shaped.ai/client');

async function retrieveRankResults() {
  const client = shapedai.Client('your_api_key');
  const model_name = 'hompage_search_and_recommendations';
  const user_id = '1';
  const config = { limit: 5 };
  const results = await client.rank({ model_name, user_id, config });
  console.log(results);
}

For the Product Manager: A Growth Engine vs. a Search Utility

For a Product Manager, the choice is between a tool that solves a feature request and a platform that drives core business KPIs. Algolia is an excellent tool for implementing a great search experience. Shaped is a growth engine designed to improve engagement, conversion, and retention across every user touchpoint.

Product Manager Perspective: At a Glance

A Platform Built for the Whole Product Team

Algolia is powerful, but many business-level changes require engineering support. Shaped is designed to empower both sides of the house. We provide a user-friendly dashboard where you can create and manage business rules. Need to launch a promotional campaign and pin specific items? You can do that with a few clicks, without waiting for an engineering cycle. This means you can move with agility while your engineers focus on the core models.

Making the Right Choice for Your Team

The Trade-offs: When is Algolia the Better Choice?

If your problem is 100% front-end keyword search with no plans for deep personalization: and your product roadmap does not include surfaces like dynamic feeds or personalized recommendations, Algolia's laser focus and InstantSearch libraries are purpose-built and highly optimized for this task.

By being transparent about this trade-off, we hope to help you make the best possible decision for your specific context.

A Final Note from the Founders,

We built Shaped because we were the MLEs and relevance engineers stuck with the limitations of traditional systems. We spent countless hours writing brittle 'glue code', fighting with data pipelines, and trying to peer inside black-box APIs. We built the platform we always wished we had—one that is transparent, powerful, and a joy to use.

We believe the best builders deserve the best tools. We hope you see that philosophy reflected in our product and our approach.

We look forward to building with you.

‍Ready to try Shaped? Schedule a deep dive with our founding engineers. We'll whiteboard your specific use case and show you exactly how we can get you up and running fast.