Analytics Development


Analytics Portfolio Management.

Over time, organizations create or acquire analytics. In a large government organization, it is not uncommon to have thousands of analytics. Axiologic Solutions offers a set of services to effectively manage these analytic portfolios:

  • Understanding the current state of analytics
  • Deriving strategies for avoiding obsolescence in analytics
  • Providing the visibility into the totality of available analytics

The basic concept behind analytics portfolio management is to make analytics lifecycle decisions (or change management) at the portfolio aggregate level (top-down or the forest level), and not at the individual (numerous) analytics level (bottom-up or the tree level). This top-down (forest perspective) management approach is what ensures that the analytics as a whole support the business needs of the organization by helping to reduce cost and risk, providing rationalization for investment decisions, providing standardization of architectures, and facilitating a modernization/optimization roadmap.

Doing ML at Enterprise Scale (MLDevSecOps, MLSecOps).

DevSecOps is the de-facto standard for performing the end-to-end, traditional software development lifecycle. DataOps is emerging as a series of similar practices for supporting data components through their lifecycle.

The lifecycle of ML models used in analytics is sufficiently different from traditional software and data that a series of practices to support the ML models throughout their development lifecycle is required.

The Axiologic Solutions MLDevSecOps approach describes how to apply DevSecOps principles (as well as others) to automate all aspects of the ML model lifecycle, including feature preparation, model training, model validation, model selection, model packaging, model deployment, model monitoring and model re-training.

The Axiologic Solutions MLDevSecOps adopts traditional DevSecOps practices and applies them to the needs of the ML model lifecycle. This includes:

  • Continuous Testing of the entire ML model pipeline; the quality of the training data; the quality of any model.
  • Continuous Integration of data; features; different software used in the pipeline; integration of models into the pipeline; integration of models into model-serving services; integration of models/services with business applications/analytics.
  • Continuous Delivery of pipelines; models; model-serving services; supporting documentation.
  • Continuous Deployment of data pipelines; model pipelines; outputted models in a variety of locations.
  • Continuous Training of models and support of different training/learning modes such as on-line training, incremental training, batch training, and federated training.
  • Continuous Monitoring of the accuracy of the model and the context of how it is being used (through a model-serving service or other method).
  • Continuous Learning of the context of using models, observing feedback and side-effects.

The Axiologic Solutions MLDevSecOps seeks to add maturity, consistency, automation, and standardization to the entire ML model lifecycle by defining processes and techniques to make the ML model lifecycle more reliable, predictable, and productive.

Business Intelligence.

BI comprises a methodology and technologies used by enterprises for the analysis of data to support a wide range of business operational decisions. Typically, BI focuses on data analysis for understanding operations; data analysis for mission support is a typically covered by the field of data analytics.

We have created a “modern” approach to BI, with the following key features:

  • Multi-dimensional agility
  • Tight integration with other teams/functions
  • Emphasis on visualization
  • Use of machine learning techniques
  • Support for emerging data domains
  • Prescriptiveness of overall approach