Amar Parsekar

• Optimized Spark on EMR (partitioning, shuffle tuning, skew mitigation, caching) for large-scale distributed workloads
• Improved performance via executor sizing, memory tuning, and parallelism optimization, reducing end-to-end job latency
• Designed cost-efficient EMR clusters (Spot instances, auto-scaling, right-sizing) for high-throughput pipelines
• Debugged and stabilized shuffle-heavy pipelines (skewed joins, resource bottlenecks, failures) improving reliability in production

GenAI, Agentic AI

Leading development of GenAI infrastructure and distributed inference systems for large scale data processing.

Key Contributions:

• Architected GPU-based LLM inference platform capable of handling thousands of requests
• Built distributed AI processing pipelines using Ray enabling parallel processing of large datasets
• Developed LLM powered systems for business entity analysis and website intelligence
• Designed scalable API-based LLM inference services deployed on AWS infrastructure
• Implemented high-performance pipelines using asynchronous processing and batch inference
• Designed production-grade monitoring and observability for AI systems

Tech Stack: Python, Ray, AWS, GPU inference, distributed systems, LLMs

• Successfully scaled app from 100/s to 1000/s by identifying bottlenecks and implementing optimization strategies with AWS cloud and Auto Scaling Group (ASG)
• Implemented best practices for machine learning operations (MLops) in data science environment, including infrastructure setup using Ray Framework and Docker, automation, and collaboration
• Implementing ChatOps strategies to facilitate AWS monitoring and improve overall efficiency in cloud infrastructure management.
• Tech stack - Vert.x, Apache Kafka, Kafka Streams, Amazon Web Services (AWS), Java, AngularJS, Spring Framework, Python

• Successfully migrated current version control tool from SVN to Git, including setup, migration and training team members.
• Assisted in implementing CICD pipeline in the project, including setting up automation tools and integrating with other systems.
• Created data warehouse to collect data from various systems to analyze patterns in security lending and borrowing utilizing SQL, including data cleaning, loading, and creating reports and visualizations.
• Tech stack - SQL, Shell script, Java, Python, Tableau, AngularJs

• Understanding structure of user queries coming from bot, processing those queries and making them ready for training
• Understanding the flow of the conversation, and differentiating between Good and destructive paths
• Integrating AI systems with Enterprise applications
• Understanding user stories and doing development by adhering to the SDLC cycle
• Tech stack - Python, NLP, Data Analysis, Data cleaning, Java, AngularJs, Apache camel

• Developed application from scratch utilizing Java as backend and AngularJs as frontend, resulting in a robust and high-performing application.
• Implemented a cost-effective custom XML based PDF generator framework, increasing efficiency and reducing costs.
• Contributed to the development of features such as Nominee generation and Account statements, improving the application's functionality and usability.
• Tech stack - Java 8, Spring 4, Hibernate 4, JIRA, Angular Js, Backbase, GitHub, Apache camel

• To Develop new transactions based on FRD
• To fix defects encountered in SIT/UAT
• To prepare test cases for Unit testing
• Discussing functional requirements with BA/Requirement team
• Doing deployment using Jenkins, finding possible resolutions in case of exceptions
• Giving testing team walkthrough of newly developed functionality
• Preparing Change documents for Enhancements
• Tech stack - Java 7/8, Spring 4, Hibernate 4, Service Now, GitHub