# Compute Infrastructure - Efficiency, Autoscaling and Offline Workload scheduling

## What am I working on currently?

Building platform to run offline batch jobs on K8s, smarter cluster selection strategies, faster scheduling and migration of customer to the consolidates platform.

Led the development of an automated experimentation framework to optimize compute and memory utilization across thousands of production workloads, improving workload stacking and resource utilization and driving sustained efficiency improvements at scale.

Contributed to build and run a predictive autoscaling pipeline for stateless Kubernetes fleets, combining traffic validation with resource modeling to make intelligent scaling decisions.

Drove cost-tracking initiatives by integrating resource usage with hardware forecasting, enabling chargeback-style analysis and fostering a resource-aware engineering culture in a non-cloud environment.

Championed efficiency engineering best practices, helping service owners reclaim unused resources and unlock meaningful, measurable cost savings.

# Performance Engineering at Linkedin

Developed and maintained client-side instrumentation libraries for real-time user monitoring (RUM), tracking page and app load times on web and mobile.

Enhanced member experience through personalization, performance improvements, and detailed user experience tracking.

Conducted large-scale data analysis to generate actionable insights by correlating key metrics like QPS, session growth, and weekly active users (WAU) with real usage data, enabling partner teams to make informed decisions on data center usage, licensing, and financial planning.

Built low-latency data pipelines powering ML models that predict network quality in real time.

Journal Technologies supplies case management software to courts, prosecutors, public defenders, probation and other justice agencies.

Worked on eCourt's technology stack ( Struts, Bootstrap, React ) & completed more than 30 tickets. Familiarized with dev tools like Git, Confluence, JIRA, Bamboo, Docker.

Worked on the 'Timeslot View' for judges to categorize case events & enhance the usability of case summary dashboard.

Understood the HazelCast's cluster level cache implementation & performed load testing on the clusters using jMeter to find upper cap.

Secured the raw entries and API keys , kept on servers in DMZ zone by integrating HashiCorp's Vault & two-factor authorization.

Upgraded the case filing portal (PHP/ Drupal) to higher versions & integrated existing payment gateway. Wrote Groovy scripts to embed business rules for fee calculations.