◦ Led the design & re-architecture of a heavily used (1M code submissions daily) monolithic code-checker service to a distributed container (Runc, Kubernetes) based run-time using Golang. This in turn improved auto-scaling of the code-checker by 80% and reduced the time to add/upgrade a language runtime by over 90%.
◦ Launched several highly available (99.99% uptime) components and features for the code-checker service like custom-checker allowing users to customize scoring process using custom code snippets.
◦ Added custom metrics using prometheus in code-checker and integrated with stack-driver for alerting, logging.
◦ Saved 30% in cloud infrastructure costs by moving less critical metrics to a self-hosted prometheus server.
◦ Led the engineering effort to improve failure rate for a critical pipeline responsible for processing over 800k HackerRank test attempts monthly which reduced overall failure rates by 70%.
◦ Developed a ReactJS wrapper for the monaco editor that supports LSP based auto-completions and migrated the HackerRank’s code editor from CodeMirror to Monaco.
◦ Decoupled the kubernetes infrastructure for Ruby on Rails based Candidate-site from HRW enabling them to scale independently.
◦ Created geo-distributed kubernetes clusters for the Autocomplete service to attain sub 100ms latencies.
