Karam Jyot Singh Gujral

Experience

• 

  Cognizant

  Oct 2014 - Dec 2017

  Programmer analyst

  Client Name: E*TRADE (subsidiary of Morgan Stanley)• Implemented various Oracle enhancement in-memory column store, for better query performance. • Implemented editioning in Oracle database having zero downtime for database updates. • Designed and Developed Shell script pipeline for deployment of database structures from git updates into all 30+ development servers and 16 production servers in parallel with fully automated pre and post validations. • Created custom revenue reports and analysis of key metrics using Tableau. Show less

• 

  Verizon

  Dec 2017 - Nov 2021

  • Designed, trained and productionized customer call Intent model and customer call propensity models in order to pre-emptively engage customers, by building multi-layered machine learning solution. Performed drift monitoring and quantitative analysis of model and its revenue impact and further enhancing the solution in next iteration to multilabel classification, ultimately helped in reducing 20k customer calls per month from customer engagement channels.• Constructed business use case around product cannibalization by performing predictive analysis. Developed and designed several proof-of-concept models using Log-log regression, Markov chain model (Bayesian Statistics), finally provided customized solutions using structural time series. Identified existing cannibalization scenarios in one of the key Verizon products using variational inference model prediction, helped in identifying potential ongoing loss of about 500K$ per month to company.• Created new promotional/marketing and customer experience ideas by identifying observational, comparative and existing predictive insights which increased revenue generation up to 300k$ per month, it helped increase CTR and Conversion rate for primary services.• Designed central insight repo database which provided real-time insights to customer engagement systems for customer personalization and business decision making systems.• Identified behaviour cohorts related to customer fraud and developed predictive model for minimizing invoice to payment loss as a proof of concept and provided predictive insights using Benford’s law and other features which helped identified fraudulent transactions with 93% precision.• Enhanced customer experience and conversion rate by creation of effort score on website to identify customers proactively who might need guided assistance. Deployed the solution which increased conversion rate by 5-6% for specific products and contributed in improving Net promoter score. Show less • Architected and migrated on premise DB2 database to AWS Postgres, saved around 10 million$ in terms of licensing cost and server maintenance cost, used AWS schema conversion tool, AWS pipeline. • Developed custom pipeline using system triggers and procedures for fail safe backup. Developed key metrics related to database, procedures and sql performance on Kibana dashboards.• Converted 400+ procedures from DB2 to AWS postgres database and created automated system for performance monitoring comparing converted procedures in postgres with DB2 procedures, decreased average runtime of procedures more than 30mins in some cases. • Awarded Spotlight award from organization for all the above efforts. • Generated business reports and insights related to customer services, products and revenue in Oracle analytics. Show less

  • Specialist - Data science

    May 2018 - Nov 2021

  • Analyst System Dev

    Dec 2017 - May 2018
• 

  Spearline®

  Feb 2022 - Sept 2023

  Data scientist

  Spearline a Cyara owned company• Designed and trained an AI model based on combination of supervised and unsupervised algorithms to detect transcoded calls on different network providers for our clients, currently pending patent approval.• Developed a mathematically optimized scoring algorithm using various audio, video, and network features for WebRtc based product “agent assure”.• Designed and trained an AI based ranking mechanism for our customer facing platform.• Trained Attention based Graph Neural network for identifying carrier information across globe.• Identified use case while engaging stakeholders and developed an audio segmentation transformer-based model that accurately classifies different types of noise frame-by-frame.• Designed and built framework for automating critical support tasks for our clients using custom build speech transcription, audio tone identifier deep neural network and using clustering algorithms achieved 80%+ automation ,50%+ reduction in ticket resolution time and 60%+ reduction in support engagement time.• Led the HubSpot, Zabbix, Google analytics data migration for analysis, data visualization and dashboard creation. Developed ML pipelines using Airbyte and created delta lake using Dreamio , helped in setup for MLFlow for Machine learning and Deep Neural network experimentation tracking.• Currently working on developing non-intrusive voice quality AI solution using Vision transformer and improving upon existing Pesq and Polqa algorithms and their drawbacks.• Currently training internal chat bot using Huggingface / OpenAI solutions to customize for specific organization use case. Show less

• 

  Eaton

  Sept 2023 - now

  Senior Data Scientist