Gaurav Gupta Email & Phone Number

Portland, Me, Us

GENAI Productized Experience:1. Automated Unit Test-cases Generation using AWS Sonnet 3.5 and AWS OpenSearch.2. Unified Customer Data Platform -> Segmented Insights using AWS Sonnet 3.5 and Pinecone.3. Documents -> Categorized Domain Level Information using Open AI, Azure Document Intelligence, Neo4J DB, Pinecone.Working as hands-on Gen AI Leader with 15-18 engineers to setup generative AI practice for various clients. Provide expertise in Fine Tuning LLM, Creating LLM from Scratch on Local/Cloud, Advanced RAG Implementations (Astute RAG), Data Engineering & Frameworks (Data Mesh, Data Lakehouse), machine learning algorithms & solutions (Attribution Modelling), Multi-modal LLMs (Text, Image, Video) etc. Well-versed with Tools/Frameworks like AWS Bedrock, AWS Sagemaker, Crew AI, AutoGen, Azure Document Intelligence, GCP Vertex AI, Databricks, Dremio/Starburst, GraphDBs (Neo4J) etc.

San Jose, Ca, Us

• Architected & Managed team (8 engineers) in Adobe Document Cloud(PDF) to create end-to-end legal contract review solution for enterprises that include features of Q&A, Contract Summarization, Entities & Clauses Extraction, Attribution etc. using both open source and standard LLM models (GPT4/ Claude Sonet). Developed Subjective and Objective evaluation framework for benchmarking of this LLM based solution against target metrics of 90% accuracy/precision/recall. Also, tried other approaches which include RAG implementation using Azure and fine tuning LLAMA3-8B using PEFT and LORA on AWS Sagemaker (Jumpstart).• Part of core team to have "Centralized Data Management Platform (CDM)" that include key features of searching, filtering, look ahead caching, >90% GPU Utilisation for ML training, versioning within strict compliance & privacy guardrails. This CDM has been built using Databricks, Spark Streaming, Redshift, MLFLow, AWS ML Stack, MosiacML MDS Format. The Main focus of this CDM is to curate a set of high-quality, comprehensive, and ethically sourced AI training datasets to build multi-modal models – images, videos, text, vectors, 3d etc. These models have been used in Adobe Firefly & Kineto for image & video training respectively.• Architected cost-effective Deep Learning Platform for Document Cloud (PDF) for running 50+ deep learning models [TF, Pytorch] catering to PDF Parsing on AWS Stack using AWS Sagemaker, GPTJ, AWS EC2, AWS S3, Sagemaker Pipelines, AWS Glue, ECR. Reduced yearly Cost of Platform by more than 1 Million USD in less than 6 months.

Greater Noida, Up, In

# Worked in building learning management solution from scratch on AWS Infra called as "Yucampus" [https://www.yucampus.com/]. Includes basic LMS features like Quiz, Assignments, Live Teaching, Classes & Groups, Infographic resumes etc. to help students in Uzbekistan for e-learning.**This project was earlier named as"Chalkmates".

New Delhi, India, In

# Helped in developing & scaling engineering infra from scratch on AWS stack for 1:1 English teaching for Kids. Insured 99.99% service availability.# Helped in developing unique video software to simulate LIVE English teacher using video sequencing.# Helped in developing AI software to understand and improve English with Virtual tutor.

Gurgaon, Haryana, In

# Worked in developing Machine Learning Model (Accuracy : 92%) for creating diet plan on people habits & preferences.# Worked on guiding platform security, application development & mobile app development.

Florham Park, New Jersey , Us

#Hired/Lead team of 14 engineers to develop insights pipelines for Verizon Network for 3 different pods: Network Transparency, Ticketing and Change Management.#End-to-End Pipeline migration from Yahoo Grid to Edge AI Infrastructure (Kafka, Java, Looker) for Verizon within 3 months of hard deadline.# Developed MVP for Customers Knowledge Graph for PNC bank using Neo4J, AWS Stack, Cipher Query.

# Developed end to end "Customer Data Platform(CDP)" with key features of UCP(Unified Customer Profile), external data integrators (Salesforce, HubSpot etc.), Text2SQL model, Outbound (Email) channel. # Developed 4 core ML models i.e. Churn Prediction, Lead Scoring, Upsell/Cross-sell & RFM Segmentation.# Developed End to End Email orchestration using Unlayer & AWS SES for sending offer emails for targeted customers with >90% delivery rate.# Developed unique "Econometrics model" for model result correlation with market sentiments for actionable insights.

Toronto , Ontario, Ca

Blockchain and AI driven open marketplace to provide every individual, enterprise, non-profit organizations & research universities a platform to monetize & exchange data in a regulated economy. Mainly created to exchange medical health data to spearhead Cancer and Chronic care research for betterment of society. Backed by one of the largest incubator in Canada i.e. Toronto Business Development Center. # Contributed in COVID-19 by predicting healthcare resources using NVIDIA Jetson Nano.# Helped AIIMS and IIT-Delhi join venture in spearheading Aortic Heart Attack prediction by analyzing thousands of full scan MRI images (DICOM Format).**Unable to pursue further due to COVID impact on CANADA immigration.

New Delhi, Delhi, In

# Develop machine learning model to predict mobile app for acquisition using various factors such DAU, WAU, MAU, Developer profile, reviews etc.# Develop initial CDP model for incoming millions of Data signals using AWS Cloud Stack.

# Life Member (LM- 00756) of institute to participate in reforming Indian democratic values.

San Jose, Ca, Us

• Built Personalization platform (A/B Experimentation testing) from scratch to test various statistical hypotheses using both traditional Control-variant population and scientific multi-arm bandit algorithms for omni-channels communication (Email, Notifications, In-App). And also, built multi-channel attribution algorithm using Shapley Value for campaign effectiveness and eventual production roll-out.• Architected and Development of near real-time analytical & custom-metrics pipelines for "Adobe Personalization platform named "Sophia"" to measure business metrics (like conversion or retention or engagement) or surface metrics (like Email Open rate, bounce rate etc.) using Hadoop (Hive Queries), Java, AWS Cloud Stack (Kinesis/S3/Dynamo DB) which has increased campaign evaluation effectiveness by 212%.• Architecting and Development of "Dynamic Content Provider (DCP) Service framework" using "Java/Spring Boot/REST Services" for self-service integration of any Machine learning model into personalization platform resulting in increase in user's engagement by 33%, conversion by 17% and retention by 13%.• Orchestration of Personalized Email System using Java/Spring Boot, Adobe Campaign, Apache AMQ, AWS Cloud Stack (EC2, SQS) to send ~50 million monthly emails within 99.99% SLA. Implemented various algorithms such as Email Fatigue, Email Conversion Matrix etc. for highly effective email engagement rate.• Development of various Machine learning algorithms like Campaign attribution, Churn Prediction, Offer Recommendation for Adobe.com, Tutorial Recommendation for Adobe Photoshop etc. using Pyspark to enhance on average user's engagement by 17-18%, conversion by 10-11%

San Jose, Ca, Us

•Involved in Development of "Customer Data Platform" to create "Ideal Customer Profile(ICP)" by extracting and analyzing billions of Adobe customer's profile/product-usage/marketing/sales-data [ranging in PBs] coming from various ETL sources like Hadoop via Apache Kafka, AWS S3, Adobe Analytics, salesforce.com (sales data) & other offline sources to provide actionable insights (in form of creating business funnel, knowledge graph, Sequencing) using Pyspark, Hive Queries which has resulted in improving customer's NPS score by 43%.•Involved in defining and measuring Key KPI metrics for measuring new product for e.g., Project Nimbus or new acquisition for e.g., Adobe acquisition of Stock or Behance or new feature for e.g., CC Libraries integration into Photoshop using Adobe Analytics/Tableau/Hadoop.•Involved in development of Text mining NLP product application titled "Manthan" [Integrated within other Adobe products] which include features like text-summarization, text-categorization etc. using Java/Stanford NLP/Python which has resulted in average increase in user's engagement per product by 14-18%.

Delhi, Delhi, In

# Taught "Management Information System(MIS)" to full time MBA students with FMS dean Mr. ML Singla.

New Delhi, In

# Help in doing Social media analytics and competitive analysis thereby boosting sales by 80% in just 3 months.# Help in driving email and notifications campaigns for off season promotions to boost sales by 75%.

Amethi, Uttar Pradesh, In

# Taught Business Analytics & Machine Learning to MBA students

Chennai, Tamil Nadu, In

# Taught "Data Visualization using Tableau" to Executive MBA experienced professionals.# Taught "Business Analytics" to full-time MBA students.

Noida, Uttar Pradesh, In

Development of 3 financial Add-ins namely i.e., Account Receivable, Sales Schedule, and Data Archival using C#.Net, SQL Server, ADO.Net for accounting features integration [with QuickBooks] which has resulted in sale of product in US & Australia by 8%.

Noida, Uttar Pradesh, In

Independently developed “A Java based Multithreaded simulator used to replicate Online Stock Exchange Functionalities like Dummy Market, Dummy users, Dummy trading etc." to reduce failure rate of Live exchange by .1%. [This has been used with 6+ stock exchanges including Singapore Stock Exchange (SGX), OMX (Sweden), ASX (Australia)]

Noida, Uttar Pradesh, In

Developed activity kit to help 10th /12th class school students to measure & control common physical parameters like temperature, light or sound intensity, relative humidity etc. using a commonly available computer using 8051 Micro- controller programming, Java, HTML5. [Ref: https://vigyanprasar.gov.in/science-communication-programs/activity-kits/pc- interface-for-science-experiments/ ]

Executive Mba, Marketing & Strategy

B.Tech, Electronics And Communication

Non-Medical, Science