The bank is designing a
Speech to Text system to personalize the IVRS options for customer complaints. To identify the
appropriate options, LDA topic model was run on the past customer written complaints. This is also being
used to predict the IVRS option (topic) after converting the Customer’s voice to text as part of the Deep
Speech language modelling pipeline.

The bank is designing a Speech to Text
system to personalize the IVRS options for customer complaints. To identify the appropriate options, LDA
topic model was run on the past customer written complaints. This is also being used to predict the IVRS
option (topic) after converting the Customer’s voice to text as part of the Deep Speech language
modelling pipeline

The bank is designing a Speech to Text
system to personalize the IVRS options for customer complaints. To identify the appropriate options,
LDA topic model was run on the past customer written complaints. This is also being used to predict
the IVRS option (topic) after converting the Customer’s voice to text as part of the Deep Speech
language modelling pipeline

APAC cards team wanted a more personalized
recommendation setup for retail customers to boost card usage. I proposed and designed a system
which could leverage the sequence of cards swipe or online usage at merchants and partners. This
setup learnt the similarities between various merchants/product c categories based on the context
in which they have occurred. Word2Vec concept was utilized to send recommendations to
customers on the basis of recent transactions (within 30 days). This also provided scope for
recommending a new personalized card types with appropriate offers on similar merchants. I also
used Locality Sensitive Hashing for fast and real time calculation of similarities