Dhruv Varshney

Engineered a production-grade pipeline for translating 300+ page medical documents, using Python, Flask, Google Cloud APIs, and custom parsing systems. Developed both backend and internal tools with full debugging, testing, and performance optimization. Built a modular microservice-style DOCX reconstruction engine ensuring consistent formatting across large file workflows. Implemented high-reliability workflows under compute and cost constraints (reduced GCP spend to < $300).

Designed and implemented backend components for a multimodal ML system using Python, FastAPI, and custom inference pipelines. Built and debugged end-to-end workflows using Apache Airflow: ingestion to processing to merging to retraining. Improved reliability of distributed jobs using Redis queues and containerized deployments in Docker. Implemented automated versioning, testing, and rollback mechanisms for production ML systems. Integrated PostgreSQL-based metadata tracking and MLflow for experiment logging.

Built and deployed serverless backend APIs using AWS Lambda and API Gateway, handling auth, routing, and DynamoDB integrations. Assisted in designing distributed workflows and event-driven systems using S3 triggers and CloudWatch. Managed IAM policies, logging, monitoring, and performance tuning for production applications. Led a 5-member team in building scalable HRMS and IT recommendation engines on AWS.