Aug 2021 – Dec 2023 · 2 yrs 4 mos · San Francisco Bay Area

• Worked on the LLM platform enablement and vendor evaluations (for vector stores, prompting, indexing, quantization and GPU footprint reduction) for NLP based (transcript/document summarization) Generative AI use cases
• Worked on real-time models (Seldon-core), batch models, feature store and other use-cases for Verizon Business Group
• Worked on Computer Vision platform proof of concept - Image DB, Image Labeling and other data prep components
• Worked on the API Gateway (APIGEE), docker image release pipeline, benchmarking for pose detection (CV) models
• Enabled distributed computing through Ray (for spark and DL use-cases) for various models
• Worked on the end to end lifecycle of various types of ML models for Google Cloud Platform based hybrid cloud MLOps platform
• Migrated ML platform components from Yahoo to Verizon and worked on their stability/optimization

Feb 2021 – Aug 2021 · 6 mos · San Francisco Bay Area

• Worked in Big-data Machine Learning group at the intersection of software engineering and machine learning.
• Implemented tier-based CPU compute resource allocation and eviction to handle underutilization for on-premise hosted Jupyter notebook service platform (like “Google Colaboratory” on premise)
• Optimized the ML platform build release (CI/CD) process causing US-West cluster failure (build storm) via a step by step hierarchical release (from an exponential to a linear count - to a best to at most 3 times) for ML docker images’ build process in the ML-Bundle pipeline
• Optimizing ML image dependency graph
• Streamlined tier-based GPU compute resource allocation and eviction to handle underutilization for on-premise hosted Jupyter notebook service platform (like “Google Colaboratory” on premise)
• Added fixes to internal model/feature store platform
• Fixed various ML sample/starter notebooks for data scientists’ productivity

Jan 2020 – Feb 2021 · 1 yr 1 mo · San Francisco Bay Area

• Data Science track lead for Verizon Telco Marketflow project
• ISP labeling using household graph for 150M unique users (90M unique households)
• Improving Data quality for the projects and KPIs created across data from 2 years, long tail data included
• Removing data bias from certain carriers
• Data/ETL Pipelines created/managed for project
• Internal paper submission : "Political Insights: the understanding of political leanings and targeting of political campaigns"
• External talk : "Recommendations in the wild"

Apr 2019 – Jan 2020 · 9 mos · San Francisco Bay Area

• Working in Insights group under Data org in the following teams :
• 1. Audience Insights
• 2. Consumer Insights
• 3. Politics Insights
• Publications:
• 1. Defensive Publication : “Method and System for Conducting Online Political Polling Using Machine Learning Models” IP.com, 2019 (Details under publications)
• Submissions:
• 1. “Politically Sensitive: Poll without Asking”, Yahoo! (VerizonMedia) TechPulse, 2019
• 2. “Anomaly Detection in Online User Activity”, Yahoo! (VerizonMedia) TechPulse, 2019
• 3. "Quantifying content trends and approximating KPIs", Yahoo! (VerizonMedia) TechPulse, 2019
• Data Engineering and Science projects:
• 1. Building and productionizing end to end big data pipeline for Consumer Insights product
• 2. Building CI/CD pipeline for staging for Audience Insights product
• 3. Brainstorming and Ideation process for data science features in Politics Insights product
• 4. Maintenance and debugging production/customer issues
• 5. Partly working on productionizing clustering pipeline
• 6. Adding data (ETL) related features to the product
• 7. Generating recommendations for user segments using Collaborative Filtering (In Progress)

May 2018 – Mar 2019 · 10 mos · San Francisco Bay Area

• Part of a selective 2 year rotational program for engineers.
• Currently working in Insights group as a part of the first yearly rotation.
• Teams (Data Org.):
• 1. Audience Insights (AI)
• 2. Consumer Insights (CI)
• 3. Politics Insights (PI)
• Submissions:
• “Improvements to Audience Insights : Top Insights and Recommendations”, Yahoo! (Oath) TechPulse, 2018
• Projects/Experiments (check 'Projects' section for details):
• 1. Improving Audience comparison metric (affinity) using distance/correlation measures (AI)
• 2. Audience segment/cohort understanding and deriving key distinguishing factors using Clustering (AI) (partial)
• 3. Anomaly detection in Counters data for proactive prevention of data indexing pipeline failure (AI)
• 4. Audience and item group recommendation using purchase funnel modeling for better/selective segment creation using collaborative filtering/recommender system (AI) (Ideation)
• 5. Ranking for trending content titles and topics for various content properties across the company (News, Sports, Finance etc.) (CI)
• 6. Sentiment analysis for content based on text in articles for various content properties across the company (News, Sports, Finance etc.) (CI)
• 7. Political bias/lean score for news content domains (PI)
• 8. User partisan lean score calculation across political parties (PI)

Feb 2017 – Apr 2018 · 1 yr 2 mos · San Francisco Bay Area

• Teams (Cloud Services, Science and Technology Org.):
• 1. Public Cloud Platforms (PCP)
• 2. Platform as a Service (PaaS)
• 3. Infrastructure as a Service (IaaS)
• 4. Yahoo Query Language (YQL)
• Publications:
• “Auto-scaling for Hybrid Cloud based Apps : Reactive, Predictive and Hybrid,” Yahoo! (Oath) TechPulse, 2017
• Projects (check 'Projects' section for details):
• 1. Logging solutions for Kubernetes on AWS (PCP)
• 2. Moving YQL tables from Sherpa (Yahoo’s distributed NoSQL key-value store) to S3/DynamoDB on AWS with data encryption using AWS KMS (YQL + PCP)
• 3. Certificate management, mutual authentication (TLS) and service creation (YQL + PCP)
• 4. Reactive and Predictive Auto-scaling for apps on docker container based PaaS platform (YQL + PaaS)
• 5. Simple Node.js app for Cloud.corp (a platform for onboarding beginners at Yahoo) with extensive documentation​ (IaaS)
• 6. Migration of legacy PHP app in production to docker container based PaaS platform (YQL + PaaS)

Jun 2016 – Sep 2016 · 3 mos · Sunnyvale, California

• As a member of Platform as a Service team in Cloud Services group :
• Learnt about CI/CD model and multiple things about cloud, application deployment, build configuration and large scale distributed systems
• Worked on configuration addition in servlet containers for upgrade of the internal cloud
• Worked on optimization of image layering in docker to reduce build time for deployed applications in YQL
• Worked on design and architecture of system for reducing pull time from docker registry for large clusters
• Technologies :­ Perl, Git, Docker, Python, Jenkins

Jul 2014 – Mar 2015 · 8 mos · Gurugram, Haryana, India

• Worked on a randomized algorithm for simulating Stock Exchange system based on several parameters
• Worked on the transaction approval process for a Master Data Management System for trading centers in different parts of world
• Introduced to technologies such as Java, J2EE, Hibernate Framework, Spring Framework, Spring MVC Framework, Apache Tomcat server management, Maven

Jun 2013 – Aug 2013 · 2 mos · Greater Delhi Area

• Worked as an intern/trainee in the Hydrophones Group of Solid State Physics Laboratory(SSPL), Defence Research and Development Organization(DRDO) - Study on Autonomous Underwater Vehicle as a sensor

Dec 2012 – Jan 2013 · 1 mo · Ahmedabad Area, Gujrat, India

• Autonomous Underwater Vehicle Systems Development

Sep 2012 – Jul 2013 · 10 mos · Greater Delhi Area

• Led a team of 22 students; Team Seaboltz, an initiative in the field of naval research.
• Was actively involved in Electronics Design, Acoustics Design, Computer Vision and Market Research
• Role involved effective Team Management; tasks included sensor usage, design of Passive SONAR system using omnidirectional hydrophones, Vision Module creation
• Project covered by CNN-IBN news as part of innovations happening across Netaji Subhas Institute of Technology

Jul 2011 – Aug 2011 · 1 mo · Greater Delhi Area

• Worked as an intern/trainee in the Network Services Division(NSD) of Defence Scientific Information & Documentation Center (DESIDOC), Defence Research and Development Organization(DRDO) - Study on Topologies of Complex and Hybrid Networks