Hi, I'm
Harshit Kumar Gupta
Machine Learning Engineer at GenAI Innovation Center AWS
A versatile Machine Learning professional building scalable and responsible AI Applications using ML models and evaluating them on various KPIs, then deploying as microservices. Passionate about learning new technologies and mastering them in a very short time span.
Currently leading ML-powered anomaly detection systems at Amazon, processing High volume of Selling Partner Payouts.
Machine Learning Expert with specialization in Computer Vision, NLP, Deep Learning, and Large Language Model.
Current Tech Stack: Python, Java, AWS Lambda, DynamoDB, Apache Spark, Docker
10+ years across Amazon, Loyalty Platforms, CognitiveScale, Intel Security, and Cadence Design Systems.
Featured Works
Here are some of my most impactful projects across Amazon, Loyalty platforms, and AI systems, showcasing expertise in distributed systems, machine learning, and large-scale data processing.
BSOR GenAI Agent
Entry point and orchestrator for all GenAI use cases for Selling Partner Event Aggregation- Simplified Onboarding Migration for already onboarded events, E2E testing across multiple Services, Customer Asks and Operations.
Key Achievements
- • Building unified knowledge base and Retrieval system using RAG
- • Creating MCP Gateway service for existing Coral based Services(4+)
- • Creating MCP server for creating prompts and tools for SO Migration and E2E testing etc.
Transactional Deferral Anomaly Detection
Real-time anomaly detection system to detect incorrect deferral type (computed by SPYDER withholding engine) for Selling partner financial transactions.
Key Achievements
- • Designing guardrail to check incorrectly released or deferred transactions
- • Utilized event features from Seller Event Aggregator and Deferral Policies
- • Built end-to-end ML pipeline with SageMaker training, inference endpoints
Intelligent Payment Circuit Breaker System for Real-Time Payment Anomaly Detection and Mitigation
Intelligent Request Gating and Anomaly Detection of Seller Disbursement handling High volumne of Monthly Disbursements using ML models for failure prediction, pending anomaly detection, bank grouping, and result code clustering
Key Achievements
- • Designed scalable system predicting anomaly detection for Selling Partner Disbursements
- • Training machine learning models for predicting anomalous scores based on past patterns
- • Built Bookkeeper to keep track of failures at various indexes
OptimalCharge: Spatio-Temporal ML for Predictive Selling Partner Collections
ML-powered charge timing optimization system for Amazon Seller Debt Manager (SDM) to improve debt recovery rates and charge success rates through intelligent retry scheduling
Key Achievements
- • Designed statistical voting classifier ensemble model to predict optimal charge retry times for failed transactions
- • Improved overall charge success rate from 40% baseline using temporal and regional features (Region, Card Type, Hour, Day)
- • Built end-to-end ML pipeline with SageMaker training, inference endpoints, and Lambda integration
Gravty (SaaS for Customer Engagement and Loyalty Management)
Designing offers and campaigns for customer attraction with cashback and points management. Customer targeting using on-fly offer creation and dynamic rules. Building Predictive and analytics model for customer, location retain and product, offer recommendation.
Key Achievements
- • Designed scalable transaction layer for realtime and batch processing
- • Designed Drools based execution engine to execute custom rules created from blockly
- • Designed Incremental Data Processing Pipeline in Data Lake
Debt Risk Hotspots ML Model
Discover patterns of features of accounts that are conducive to bad debt.
Key Achievements
- • Interpreting model to understand similar bad debt accounts
- • Defining similarities of accounts quantitatively by using feature importance
- • Stratification to understand characteristics of clusters
Patient Scheduling Constraint Optimizer
Recommend ideal patient schedules based on patient preferences and optimal use of resources.
Key Achievements
- • Building a Patient Scheduling model to efficiently schedule medical appointments
- • Using patient preferences, medical rules, time slots and facility availability as constraints
- • Solving complex combinatorial optimization using OR-Tools
Bad Debt Risk Advisor ML Model
Building predictive model to determine bad debt for Hospital Billing and Professional Billing
Key Achievements
- • Building a model for predicting bad debt accounts
- • Developing KPI for measuring model performance
- • Building jobs for distributed data processing in Hive Using Spark
Predictive User Engagement on Tax Filing System
Improve user engagement and increase Tax Filing by using the ML model.
Key Achievements
- • Building model to handle historical data and click through data both
- • Building a predictive model for user intervention
- • Improving model response time in Production to handle peak season load
Technologies & Tools I Work With
Professional Experience
My journey through the tech industry in field of machine learning, applied science and AI engineering, across leading companies in AI and enterprise software.
Machine Learning Engineer II
Leading development of intelligent disbursement systems and anomaly detection for seller payments failures.
Key Projects
BSOR GenAI Agent
Entry point and orchestrator for all GenAI use cases for Selling Partner Event Aggregation- Simplified Onboarding Migration for already onboarded events, E2E testing across multiple Services, Customer Asks and Operations.
- ● Building unified knowledge base and Retrieval system using RAG
- ● Creating MCP Gateway service for existing Coral based Services(4+)
Transactional Deferral Anomaly Detection
Real-time anomaly detection system to detect incorrect deferral type (computed by SPYDER withholding engine) for Selling partner financial transactions.
- ● Designing guardrail to check incorrectly released or deferred transactions
- ● Utilized event features from Seller Event Aggregator and Deferral Policies
Intelligent Payment Circuit Breaker System for Real-Time Payment Anomaly Detection and Mitigation
Intelligent Request Gating and Anomaly Detection of Seller Disbursement handling High volumne of Monthly Disbursements using ML models for failure prediction, pending anomaly detection, bank grouping, and result code clustering
- ● Designed scalable system predicting anomaly detection for Selling Partner Disbursements
- ● Training machine learning models for predicting anomalous scores based on past patterns
OptimalCharge: Spatio-Temporal ML for Predictive Selling Partner Collections
ML-powered charge timing optimization system for Amazon Seller Debt Manager (SDM) to improve debt recovery rates and charge success rates through intelligent retry scheduling
- ● Designed statistical voting classifier ensemble model to predict optimal charge retry times for failed transactions
- ● Improved overall charge success rate from 40% baseline using temporal and regional features (Region, Card Type, Hour, Day)
Software Engineer II
Designed scalable systems for journal processing and AWS native service migration with focus on distributed data processing.
Key Projects
SPURSH
Aggregates transaction details into journal entries that are posted to the General Ledger
- ● Designed scalable system for JournalPostmanService and JournalPostmanPreprocessor
- ● Removed Journal status publishing and ticketing dependency from OFA
Product Engineering Architect
Architected cloud platform for customer engagement and loyalty management with predictive analytics, building ML models for customer retention and product recommendation.
Key Projects
Gravty (SaaS for Customer Engagement and Loyalty Management)
Designing offers and campaigns for customer attraction with cashback and points management. Customer targeting using on-fly offer creation and dynamic rules. Building Predictive and analytics model for customer, location retain and product, offer recommendation.
- ● Designed scalable transaction layer for realtime and batch processing
- ● Designed Drools based execution engine to execute custom rules created from blockly
Senior Software Development Engineer
Building scalable and responsible AI Applications using ML models and evaluating them on various KPIs, then deploying as microservices.
Key Projects
Debt Risk Hotspots ML Model
Discover patterns of features of accounts that are conducive to bad debt.
- ● Interpreting model to understand similar bad debt accounts
- ● Defining similarities of accounts quantitatively by using feature importance
Patient Scheduling Constraint Optimizer
Recommend ideal patient schedules based on patient preferences and optimal use of resources.
- ● Building a Patient Scheduling model to efficiently schedule medical appointments
- ● Using patient preferences, medical rules, time slots and facility availability as constraints
Bad Debt Risk Advisor ML Model
Building predictive model to determine bad debt for Hospital Billing and Professional Billing
- ● Building a model for predicting bad debt accounts
- ● Developing KPI for measuring model performance
Predictive User Engagement on Tax Filing System
Improve user engagement and increase Tax Filing by using the ML model.
- ● Building model to handle historical data and click through data both
- ● Building a predictive model for user intervention
Senior Software Development Engineer
Developed EDM (Enterprise Data Management) solutions for collaborative library and design data management.
Key Projects
Allegro EDM Solutions
Collaborative Library and design data management system
- ● Involved in development of EDM (Enterprise Data Management)
- ● Designed DAO layer to support SQL, NoSQL and Graph Databases
Senior Software Development Engineer
Worked on Application Control and Change Control with Global Threat Intelligence systems.
Key Projects
Application Control and Change Control
Global Threat Intelligence system for application security
- ● Drove several features independently and contributed to end-to-end delivery
- ● Led feature to support SHA-256 for Rule Groups and Policy Discovery
Research Projects
Academic research in Machine Learning, Computer Vision, and NLP during my M.Tech at IIT Delhi and independent ML research projects with open-source contributions.
Assessment of Autism Spectrum Disorder in Toddlers using Speech Features
Designing an Android App to collect voice sample and store in the cloud
Key Contributions
- • Analysis of Speech Samples using Spectrogram and Scalogram
- • Feature Extraction using Discrete Wavelet Transform and Discrete Wavelet Packet Analysis
- • Classification of speech samples using SVM, Random Forest, HMM, CNN classifiers
- • Application of Deep Learning Convolutional Neural Network for Feature Learning and Classification
Fake Profile Detection using Machine Learning
Detect fake profiles in online social networks using multiple machine learning techniques
Key Contributions
- • Implemented Support Vector Machine, Neural Network, and Random Forest algorithms
- • Developed comprehensive fake profile detection system for social media platforms
- • Created Jupyter notebooks for interactive analysis and model comparison
- • Achieved high accuracy in distinguishing authentic vs fake social media profiles
.webp)
Twitter Sentiment Analysis using Machine Learning
Sentiment analysis of tweets using machine learning and natural language processing techniques
Key Contributions
- • Implemented Naive Bayes and SVM models for sentiment classification
- • Developed comprehensive text preprocessing pipeline with stop words removal
- • Created lexicon-based sentiment analysis using positive/negative word dictionaries
- • Built scalable sentiment prediction system for real-time Twitter data analysis
Research Specializations
Machine Learning
Signal Processing
NLP & Social Media
Open Source Contributions
Academic Projects
Course projects and academic assignments covering web development, mobile applications, security systems, and data processing during my academic journey.
Image Encryption & Decryption and Transformation
Secure & Compressed image transfer system
Key Features
- • Implemented secure image transfer with compression
Aayush
A Web platform where doctors and patients can interact, so that patients could get help online and they could find best doctors around.
Key Features
- • Built complete doctor-patient interaction platform
Weather Forecasting Application
Using Yahoo Weather API and support for Offline Queries
Key Features
- • Implemented weather forecasting with offline support
Illuminance Correction
Android App for Illuminance Correction on Image using OpenCV for Android
Key Features
- • Developed mobile app for image correction
Academic Skills Developed
Web Development
J2EE, Struts Framework, Full-stack development
Mobile Development
Android apps, OpenCV integration, Image processing
Security & Encryption
Cryptography, Secure data transfer, Compression
API Integration
External APIs, Data processing, Offline capabilities
Want to see more projects?
Recommendations & Shout Outs
What colleagues, managers, and collaborators say about working with me across Amazon, AI startups, and enterprise software companies.
I strongly recommend Harshit Gupta, who was a key contributor on my team at Amazon. Harshit consistently demonstrated deep technical strength, strong ownership, and the ability to deliver measurable business impact. He led OptimalCharge and also architected the Intelligent Payment Circuit Breaker—both of which became foundational systems for our payments platform. Harshit brings expertise in applied ML, production-scale systems, and rigorous experimentation practices. What truly stands out is his ability to translate complex business problems into elegant, scalable solutions while maintaining a clear focus on outcomes. Any team would be fortunate to have him
"Harshit has a perfect blend of engineering and machine learning expertise. I’ve seen him tackle complex problems with creativity and deliver highly effective solutions. He identified the optimal timing to charge sellers, significantly improving success rates and ensuring the sustainability of Amazon’s funds flow. He also introduced an innovative LLM-driven migration approach projected to save nearly 2,000 SDE weeks. Harshit is an exceptional problem-solver and a true asset to any team — I would highly recommend him.
"I had the privilege of working with Harshit at Amazon, and our professional relationship dates back to our college days at KNIT Sultanpur. Harshit is an exceptional software engineer who combines deep technical expertise in Machine Learning and distributed systems with outstanding problem-solving abilities. At Amazon, he led ML-powered anomaly detection systems for seller disbursements, handling massive transaction volumes with remarkable reliability and innovative approaches. He's an excellent team player and mentor who elevates everyone around him, and his ability to handle complex, high-volume systems under pressure makes him invaluable. I wholeheartedly recommend Harshit for any Senior Software Development Engineer role.
"Harshit is a very talented individual who comes up with innovative methodologies to solve ML problems. He is very good at implementing these models in live applications. He did an amazing job in enhancing risk prediction solution with significant amount of operational constraints. Very capable individual to be on a team dealing with challenging problems.
"Harshit is a delightful engineer and one of the most amicable people I've worked with. His experience in signal processing using deep learning techniques came in quite handy for our work at CognitiveScale. I must add that he singlehandedly built an entire data science pipeline capable of handling multiple hundred requests/sec traffic throughput. A great problem solver and very reliable. Any software engineering team would love having a member like Harshit on their roster.
"Harshit joined my team 1.5 years back, Harshit came with very strong fundamental and conceptual knowledge, and soon acquired good understanding of the product & process. I found him always motivated to grab the complex work and he has shown his innovative ways to simplify things. He is a problem solver and an asset to my team.
"Harshit is very committed, exhibits true enthusiasm at work and is a very good team player. He is technically very sound. He was a tremendous asset to our group and was always capable of handling multiple assignments. He is quick to understand things and has good debugging skills. He also shows sense of urgency and is able to complete his work on time. Harshit at many times has stretched to meet tight deadlines.
"🎉 Internal Shout-Outs
Recognition from Amazon colleagues for exceptional work
Udit Khimesra
Senior Software Engineer • AmazonOctober 27, 2021
Thanks Harshit for doing seamless delivery of JPPS Optimization Change. It was a complex change and required working closely with AE team.
"Professional Resume
Comprehensive overview of my qualifications, certifications, and achievements in Machine Learning, Applied Science and AI Engineering.
Technical Skills
🤖 Machine Learning & AI
📚 ML Frameworks
💻 Programming Languages
☁️ AWS Cloud
🗄️ Databases
📊 Big Data & Analytics
🚀 Specializations
Education
M.Tech in Computer Technology
Indian Institute of Technology, Delhi
GPA: 8.43/10
B.Tech in Computer Science & Engineering
Kamla Nehru Institute of Technology, Sultanpur
Grade: 81.16%
Certifications
Generative AI with Large Language Models
DeepLearning.AIMachine Learning
Stanford University, Coursera 2015Algorithms: Design and Analysis, Part 1
Stanford University, Coursera 2015Algorithms: Design and Analysis, Part 2
Stanford University, Coursera 2015Image and Video Processing
Duke University, Coursera 2015J2EE Struts with Hibernate
Professional Certification 2016Getting and Cleaning Data
Data Science Certification 2016R Programming
Statistical Computing Certification 2016Awards & Achievements
BugBash for MAC 8.0 Award - Intel Security
3rd prize winner in Mind-Hunters event, National Level Techfest, Effluence
2nd prize winner in Fill Up The Code event, Tech Carnival By Computer Society Of India
Consolation prize in Paper Presentation on "WEB 3.0", organized by I.E.I.
Consolation prize in Technical Wordsworth event organized by Computer Society of India
State Level project on Water Resource Conservation in National Children's Science Congress
Professional Qualities
Self-motivated team player with strong analytical, problem solving, planning and resource optimization skills
Possess creativity & innovation, flexibility & adaptability and interpersonal skills with leadership qualities
Passionate about learning new technologies and tools and mastering those in a very short time span
Download Full Resume
Get the complete PDF version of my resume with detailed project descriptions and technical specifications.
Download PDF Resume