• Homepage
  • Blog
  • Synerise Monad: Apply science to behavioral data. Automatically.

Synerise Monad: Apply science to behavioral data. Automatically.

10 min read
Synerise Monad - Apply science to behavioral data. Automatically.

Deploying AI effectively requires extensive data processing, maintaining separate batch and real-time data flows, and manual feature creation. Projects take months from idea to production. Learn how to apply science to behavioral data - automatically.

Focus on applications, let AI figure out your data  

Synerise Monad is a first of its kind tool for automatic feature engineering and model creation, harnessing the power of unsupervised learning. Powered by the latest research in graph neural networks and manifold learning, Monad allows Data Science teams and data-oriented power users to focus on their business applications. The rest is figured out by Monad: large and diverse datasets are automatically transformed into ready-to-use features, both for real-time inference and batch analytics.

Key expected results:

  • Significantly reduce time (even 95%) from ML idea to production
  • Eliminate effort for manual data preparation and feature creation
  • Enable all predictive modeling use cases with a single unified representation
  • Instantly use any built-in models, customize, and create others
  • Understand predictions and derive business insights from the models
  • Eliminate effort for batch and streaming dataflows development and maintenance
  • Achieve extremely high accuracy, even with an order of magnitude less training data
  • Train models on large-scale databases
  • Deploy responsible AI to your apps by explaining individual predictions with the contribution value of every row/field of every table in your database

Fixing the time-to-market of ML applications 

All customer-facing industries want to realize the tremendous power of predictive behavioral modeling, yet applications of ML take months to execute. The bulk of the effort is spent on collecting, understanding, and preparing suitable data to create meaningful features for use in predictive models – a painstaking process that can still fall short as data-rich business environments struggle to utilize numerous data sources efficiently, often resorting to using only small subsets. Hard modalities like interactions, graphs, texts, complex datatypes are typically discarded as being too difficult to analyze. As a result, real-time operations are too costly to develop and maintain for most use cases, requiring separate logical flows and infrastructure. 

Standard process of building AI application/models eliminated by Monad:

  1. Understand data
  2. Identify entities, metadata, attributes
  3. Find good joins
  4. Filter data
  5. Transform data
  6. Evaluate cardinalities
  7. Create embeddings
  8. Create feature transforms
  9. Find good parameters
  10. Identify useful features
  11. Create feature aggregates
  12. Calculate vectors
  13. Store vectors
  14. Create target variables
  15. Train models & tune models
  16. Analyze features, add more features
  17. Scale & manage infrastructure, add servers
  18. Serve models & monitor model health

How to query the future with Monad – selected problems to solve:

 

Unsupervised learning in real-time, at scale. 

Monad is able to ingest multiple large-scale data sources, both databases and real-time streams alike. In minutes, Monad performs state of the art automated feature discovery, detecting structures like graphs and hypergraphs in the data, calculating embeddings and generating appropriate data transforms for every data type. Unsupervised learning is employed to create Universal Behavioral Profiles – automatically created feature vectors representing past behavioral information about entities (e.g. customers, devices, contracts) that maximize predictive power about their future behavior. Available both in real-time and batch mode, Universal Behavioral Profiles can be used for out-of-the box model training and inference for any and all use cases critical to your business. 

Use domain expertise where it matters. 

With Monad, the time and effort usually spent on handcrafting feature creation, data pipelines, and infrastructure management can now be better utilized to choose business objectives, quickly test new hypotheses and refine modeling targets. All standard use cases like scoring customers predicting churn, identifying purchase propensities and product affinities, personalizing content and recommendations, detecting fraud and anomalies are available out-of-the-box. Users can also easily add custom features derived from deep domain expertise.

 

Empower experts.

In addition to a collection of built-in models, Monad can be used to predict any targets supplied by the user. Models can be modified, customized, and tuned for special purposes. Helpful by-products such as embeddings, entity similarities, and optimal feature transforms are exposed for power users to utilize. Monitoring changes and feature drift over time enables teams to adapt and stay ahead of ever-changing business realities and customer demands.

Let business learn from the AI, instead of teaching it.

Understanding why things happen enables us to consciously shape the future. For every model, key events and features affecting predictions can be analyzed – both in aggregate, as well as at an individual prediction level.

Primary industries:

  • Financial Services
  • Banking
  • Media & Publishing
  • Retail
  • E-commerce
  • Healthcare
  • Travel
  • Telco
  • Gaming

What You need to start?

Set of raw event-level data linked to entities being modeled (for example: customers, devices, employees, cars, players, credit cards, anonymous IDs etc.).

Example of core built-in AI applications :

  • Scoring
  • Auto-segmentation
  • Propensity/Affinity
  • Recommendations
  • Predictions
  • Real-time LTV calcs

 

Concept used by leaders
 

"We are currently using EMDE  for generating candidates to facilitate downstream recommendation systems. It generates recommendations using density-based rich customer representation. It allows us to trace customer look-alikes (‘People Like You’) to find similar users with similar cuisine/taste preferences as well as price affinity. We used Cleora for customer-restaurants graph data […] And to our delight, the embedding generation was superfast (i.e <5 minutes). For context, do remember that GraphSAGE took ~20hours for the same data in the NCR region. Cleora + EMDE gives us a generalised framework for recommendations […] We are exploring ways to use it in other applications such as search ranking, dish recommendations, etc. “ – Zomato.com Data Science team

Zomato is an multinational restaurant aggregator and food delivery company founded 2008 owned inter alia by Uber and AliPay. Zomato provides information, menus and user-reviews of restaurants as well as food delivery options from partner restaurants in select cities. The service is available in 24 countries and in more than 10,000 cities. In financial year 2021, the average monthly active users for Zomato were 32.1 million users.

 

Designed for hyperscale. Banking Sector. Case study

Challenges

Case 1: compute propensity towards products aggregated into classes. Classes represent similar financial products (e.g. investment or insurance products). The goal is to compute probability scores of purchasing at least 1 product from each group, for each customer. The predictions should reflect probability of purchase within a given timeframe from now - e.g. 2 months.

Case 2: compute propensity towards purchase via a given channel. There are multiple existing sales channels, which are grouped into digital and non-digital channels. The goal is to predict the probability of each user buying at least 1 product via the digital/non-digital channel within a set timeframe.

All models should allow to understand which factors (such as sociodemographic, seasonal or interaction-level features) motivate user decisions of purchase/lack of purchase.

Dataset and technical information

 

Effect

Synerise Monad delivers state-of-the-art models with individual propensities for 45 product / channel combinations for all 35M customers with full interpretability insights within 3 hours of pretraining and 3 hours of inference.