Nlp Engineer
CurrentCollaborating with microsoft’s IPML R&D team as a vendor, an automation task to identify abusive content in data. 400K models generated and compared in the process of automation. Classification of text(extracted from documents, emails, images, audio etc.) with respect to profanity, harassment, discrimination and threat with help of different statistical models, Machine Learning models, using services from Azure. Automating the process of generating models like, • PMI model, a statistical model classifies text with respect to the term/word frequency.• NimbusML models, Microsoft's .net machine learning framework trained for text classification. Train various models with different parameters to identify the best model to be deployed to classify abusive content.• Performing automation task with AzureML and Azure autoML to train models by performing experiments using azure cloud services with multiple processes/processors and high performance computing.• Deploying models and model generation as a service in azure with azure functions and azure app services in azure cloud and azure virtual environment. Monitoring the resource use in azure resource groups.PCI and PI compliance redaction in conversational analytics. Intent Classification for SBI Life insurance in Hindi. Capi, an n-gram weight based classification model to identify presence of policy/complaints related and regarding to product sold to customer.• PCI and PI identification, extract payment card information not limited to credit card, debit card but also passwords, OTP’s, pin numbers and personal information such as email info, addresses, contact/phone numbers etc.• Redaction of PCI and PI information extracted using techniques like Intent classification and rule/pattern matching based approach• Hindi SBI Life insurance policy, plan, Amount, nominee, etc. identification and classification of denial of information in the policy.
