"What you built in 15mins took us a year."
— Chief Data Officer | European Bank | December 2016.
Lionsgate enables enterprise data scientists to pop-up secure, distributed Apache Spark-Cassandra clusters on-prem or in the cloud.
Organisations such as IBM, Santander, Standard Life, BNP Paribas chose Lionsgate to expedite digital transformation, GDPR, BI and compliance projects.
Pop-up an enterprise grade AI architecture in 15 mins.
Lionsgate does the hard DevOps work so data scientists don't have to. Gartner estimates enterprises end up hiring three DevOps staff to support a single data scientist. Lionsgate brings that number down to 0.1.
Lionsgate Spark and PySpark environments support a huge number of analytical libraries ranging from basic stats to deep learning. For the data scientists amoungst you these include:
Random data generation.
Classification and regression:
Linear models (SVMs, logistic regression, linear regression).
Ensembles of trees (Random Forests and Gradient-Boosted Trees).
Alternating least squares (ALS).
Power iteration clustering (PIC).
Latent Dirichlet allocation (LDA).
Singular value decomposition (SVD).
Principal component analysis (PCA).
Feature extraction and transformation.
Frequent pattern mining:
Natural Language Processing.
Stochastic gradient descent.
Limited-memory BFGS (L-BFGS).
Lionsgate - Native support for distributed graph analytics:
Open source service components include TinkerPop 3.1.1 Gremlin, Titan and Spark. Abstraction layers and Lionsgate core services are proprietary Semblent IP. Features include:
Fused data models:
Allow many graphs to operate under one schema.
Distribute management system upgrades in many networks automatically.
Impose access restrictions homogeneously and obey relevant regulation with ease.
Distribute queries and aggregate results in real-time.
Relational to Graph Transforms:
Clone - copy the source a number of times.
Subset - Only include a subset of the fields.
Decision - Perform a Boolean operation on the record and either sent the record down the true or false branch.
Record Vertex - Create a vertex with the specified label for each of the records.
Record Edge - Create an edge between the vertices specified. There will be a "from" and a "to" set of vertices and the edges will be between theses vertices.
Security & Permissions.
In addition to Group-to-Service, Group-to-Company (etc) permissions, there are three other types of permissions that can be applied to data:
Model level (ArchetypeModel_permissions) a group either has access to a model or not.
Property level (ArchetypeModel_propertyPermissions) a group either has access to a property or not (if no model level access is present for a given group, property level access is redundant). These permissions are not represented to the entities themselves they're referenced when properties are read/modified within Lionsgate.
Value level ( ArchetypeModel_restrictedValues) a group has access to a specific vertex only if the value of a property meets certain criteria. If no model level, or property level access is present for a given group, value level access is redundant.
Convenient installation procedure for and configuration of Core Software Services and Lionsgate itself.
Ability to init new ComputeGroups (e.g. create a new cluster).
Ability to join new nodes to existing ComputeGroups (e.g. join a node to the cluster).
Ability to monitor individual services for individual nodes.
Configuration rewritten dynamically.
Software firewall configuration.
IMS testing framework for comprehensive integrations tests.
Per-service SSL encryption with managed security credentials as well as OS wide per-machine encryption (transparent to running services).
Data Management Features:
Ability to add global properties.
Ability to add indices which span global and local property indices.
Configs expandable for hbase, in memory, Elastic Search, Spark, Hadoop, s3.
New management schema for efficient description of compute groups, networks, applications, infrastructure and services.
Enhanced permission/authentication/encryption systems.
Management structures centralised using the system and schema root vertices.
New style for adjacency for edge source/target labels.
New global property system implemented.
Sessionless gremlin logic.
Namespacing with dot convention (modelName.propertyName).
Generic configuration manager.
In December 2016 our wings were clipped by late paying customers, byzantine enterprise on-boarding processes and downright poor product/market fit.
Right team, right product, wrong time.
We are extremely proud of what we achieved with Lionsgate; the platform is uniquely powerful. The world's first graph-compute enterprise OS - perfect for GDPR and AI.
Although it is no longer under active development we are keen to see Lionsgate put to good use. It is a powerful multi-million dollar bit of kit and a damn sight cheaper to run (TCA) than alternatives like the mighty Mesos. If you are interested in learning more about it for your project please do get in touch, we would love to hear from you.