MLOps Conference

MLOps Conference

On DataOps, productionizing ML models, and running experiments at scale.

Venkata Pingali

Venkata Pingali

@pingali

Scribble Enrich - 2nd Generation Feature Engineering Platform

Submitted May 11, 2021

Scribble Enrich - 2nd Generation Feature Engineering Platform

Majority of feature store products and discussions have a specific
pattern: Larg(ish) company, clean interfaces, and large volumes of
data. They are very impressive but not good fit for mid-market
enterprises. Scribble has built and operated feature stores for the
past few years for mid-market enterprises. That experience has led to
a different architecture and thinking around the feature stores. This
talk expands on the same.

  1. Mid-market context - Are all feature stores the same?
    (a) RoI as a first class consideration
    (b) Data quality and volume
    (c) People and skill availability
    (d) Clarity on usecases
    (e) Diversity of needs (data, integration, applications)
    (f) Time to deliver pressures
    (g) Data/system complexity limits

  2. Designing for Mid-Market - Enrich story
    (a) Fast to deploy & use
    (b) Metadata services
    (c) Optimize for people (skill, time etc.)
    (d) Flexible interfaces to allow evolution of needs
    (e) Integrate into workflows/processes
    (f) KPIs = usecases/time, cost/question
    (g) Extend upstream and downstream
    (h) Privacy as a built in capability

  3. Interoperation - How will systems emerge?
    (a) Not a binary decision, no fundamental conflict
    (b) Used in combination, on different paths, different subspaces

Comments

{{ gettext('Login to leave a comment') }}

{{ gettext('Post a comment…') }}
{{ gettext('New comment') }}
{{ formTitle }}

{{ errorMsg }}

{{ gettext('No comments posted yet') }}