Mukesh Kumar

Software Engineer with 4 years of experience designing and scaling distributed systems and AI-powered platforms at enterprise scale. Specialized in building low-latency, event-driven architectures integrating Retrieval-Augmented Generation (RAG) and semantic search. Proven track record of delivering reliable, high-performance backend systems handling large-scale document pipelines and real-time inference workloads on AWS.

Software Engineer – AI/Backend Systems - Eli Lilly - Bangalore

(2022-07)

• Designed and deployed a distributed document search platform using a Retrieval-Augmented Generation (RAG) architecture, enabling semantic search across 10+ enterprise data sources.
• Built scalable ingestion pipelines (PDF extraction, chunking, metadata enrichment) and integrated hybrid retrieval (keyword + vector search) using OpenSearch.
• Improved search relevance by 25% by optimizing chunking strategy, metadata-aware ranking, and retrieval pipeline design.
• Integrated LLM-based summarization workflows using Amazon Bedrock, reducing manual research effort by 30–40%.
• Designed event-driven architecture using SQS and AWS Lambda to process 50K+ documents/month with fault-tolerant, asynchronous workflows.
• Built low-latency REST APIs (<200ms p95) for search and retrieval services using API Gateway and Lambda.
• Implemented retry strategies, idempotent processing, and failure handling, reducing pipeline failures by 20%.
• Designed DynamoDB-based state management for tracking large-scale asynchronous workflows.
• Built infrastructure-as-code using CloudFormation and SAM, reducing deployment time by 40%.
• Developed monitoring and alerting systems using CloudWatch and SNS for ingestion pipelines.
• Orchestrated distributed ETL pipelines using AWS Glue (Spark) across cross-account environments.

M.Tech - Computer Science & Engineering - National Institute of Technology, Tiruchirappalli (2020-08 - 2022-06)