• Drove end-to-end feature development of various modules in the crypto exchange, including Portfolio Service, EarnOnCrypto, Spot Trading, Crypto Discovery, SIP, and Futures Derivatives, enhancing trading functionality for over 16M users and $150M in daily volumes.• Architected and implemented a real-time PnL service for user portfolios using a scalable, distributed worker environment with Redis Queue and DLQ support in Ruby on Rails leveraging sliding window to process 10,000 workers/minute.• Built an in-house Remote Config Service to replace Firebase, handling 200 req/sec and delivering real-time updates to 100K+ active devices, reducing cost and dependency on external services.• Diagnosed and resolved P0 issues through in-depth root cause analysis and targeted optimizations, resulting in a significant decrease in order rejections, minimized net positions in sub-accounts, optimized fund utilization and reducing monetary exposure from 2% to 0.001%.• Led a team of 5 engineers to develop a performance benchmarking platform that simulates 5x of peak loads in various patterns like burst, random, linear etc. to identify system flaws.• TechStack: Golang, Ruby On Rails, Node.js, C++, Kafka, Redis, Websockets, Kubernetes, AWS(RDS, EC2, DynamoDB).

• “Monolith to microservices” - Led the low-level design and development of the Exchange Gateway with 7 microservices handling low-latency trading data.• Designed and implemented a stateless Fetch Engine with a rate-limit orchestrator capable of ingesting 50MB/sec from 4000+ WebSocket exchange channels, boosting platform responsiveness and reliability in real-time trading.• Optimized database connection efficiency in Monolith service by implementing a custom Rails ORM patch and PGBouncer, reducing idle PostgreSQL connections by 80%, which improved resource utilization and DB performance.• Tech-stack: Golang, Ruby On Rails, Node.js, C++, Python, Kafka, Redis, Kubernetes, AWS(RDS, EC2, DynamoDB).

• Created ETL pipelines using CDC connectors and Kafka to migrate real-time data from production to a data warehouse, efficiently handling 50,000 events/min, enhancing data accuracy and reliability for Amazon’s Buy Now Pay Later feature.• Tech-stack: Python, AWS(S3, Redshift, Lambda), MySQL, MongoDB, Kafka, Debezium.

• Developed a Flask service to run a machine learning model predicting heap maps on time series and geospatial pollution data, with a Node.js service generating interactive maps, improving environmental data accessibility and visualization.• Tech-stack: Python, Node.js, Flask, PostgreSQL-GIS, Cassandra, Docker, GCP.