At Physics Wallah, I cut User LMS API P99 latency from about 3 seconds to about 210 milliseconds by restructuring MongoDB aggregation pipelines and adding Redis caching. I also traced 20–30 second API calls to collection scans across about 210K documents and rewrote the offending aggregation stages.
For the quiz and test backend, I revamped the evaluation engine to run submissions through a Redis-backed job pipeline, improving timer handling and submission correctness under concurrent load. I also designed API contracts, data models, and validation for quiz and test workflows.
I replaced cron-based polling with event-driven Redis-backed Bull queues, worker execution, and retry handling. I also added queue inspection tooling to help debug asynchronous jobs and stuck workers.
On PW's Global Content Management System, I implemented content APIs and workflows, migrated legacy SQL/RDBMS content into MongoDB, and contributed to the Skills Blog cutover from WordPress to GCMS. I’ve also contributed to a real-time voice doubt-resolution experience and introduced Jest-based unit testing and standardized CI/CD pipelines.
