Srushti Milind Jamsandekar

Engineered e-file request approval and rejection workflows for CaseHub, a Wisconsin government court case management platform, using React.js, TypeScript, Node.js, and PostgreSQL to digitize document filing across court departments. Resolved critical N+1 query patterns, inefficient joins, and poor schema design in PostgreSQL using Sequelize query optimization, eager loading, and index restructuring, cutting database load and API response times by ~30%. Refactored the AWS CodeBuild CI/CD pipeline by consolidating redundant stages, adding pre-deployment validation checks, and implementing automated rollback triggers to improve release stability and reduce deployment risk. Secured platform access for government court users by implementing OAuth 2.0, JWT-based session management, and RBAC to enforce role-level access across case workflows and filing operations.

Contributed backend features to JP Morgan's internal LLM Suite serving 230,000+ employees, using Python, FastAPI, Django, and AWS Bedrock to support document querying, research summarization, and performance review workflows. Built RAG-style retrieval pipelines using context assembly and prompt orchestration to ground LLM responses in internal knowledge bases and research documents, reducing hallucinations and improving output accuracy by ~20%. Refactored prompt templates for Claude-powered workflows to enforce structured JSON outputs for downstream LLM pipelines, and implemented token-aware chunking to optimize context window utilization and output consistency. Decomposed tightly coupled legacy backend modules into modular, reusable service layers using Python and FastAPI, reducing cross-dependency complexity and enabling faster feature development across the LLM Suite platform.

Architected the Cash Management Service and contributed to Domestic Money Transfer and Insurance services on the Vyaapaar platform, handling 100,000+ monthly transactions worth $20B+ using Node.js, Python and GraphQL. Engineered a unified KYC abstraction layer across 7 bank aggregators to standardize merchant onboarding, and implemented a CRON-based reconciliation job running every 15 minutes to resolve pending transactions across payment providers. Spearheaded the migration of a monolithic architecture into 18 independent microservices using Docker, Kubernetes, and AWS, cutting average deployment time by ~35% and improving service isolation across the Vyaapaar platform. Accelerated merchant session performance by introducing Redis caching, reducing query response times from 500ms to 320ms, and built a real-time React dashboard enabling operations teams to monitor live transactions and process refunds.