Work

Multi-tenancy and the move to GCP. I work on the Simplified Case Management (SCM) platform. I implemented multi-tenancy across it and built the SCM-Commons onboarding that cut new-tenant onboarding time by 40%. I took a mission-critical notification API up three levels on the Richardson Maturity Model without a single breaking change for its consumers, and moved data and workloads from on-prem to GCP, including onboarding a SaaS-native solution through RAMP.

From logs to pull requests. Most recently I built the Proactive Log Resolver (PLR), a pipeline that reads our GCP logs, diagnoses what went wrong, opens a pull request and alerts the reviewers. Once a reviewer approves, the release runs on its own, and every day the team gets a summary of the last 24 hours.

Tools for engineers. I also built a knowledge base and cloud skill mapper that many teams now use for cloud migration and cloud feature development. With it, an engineer fixes 5–6 bugs a day, a feature is often a single prompt, and juniors get up to speed on the team much faster. I also build LLM-powered tools that replace manual SOPs: a code-review agent, a parallel git-worktree feature processor and an integration-test agent.

Saying no with data. I ran a Bigtable proof of concept, benchmarking 500K+ records with secondary indexes. It showed Bigtable didn’t fit our relational model, which saved a costly redesign.

The job scheduler. I worked on Catalyst, Zoho’s serverless platform. I designed its distributed job scheduler on Kafka and Redis and drove it through to production: it took the shortest cron interval from an hour to a minute for every user, handled 10M+ jobs a day, and brought job dispatch latency from 50 ms to 5 ms by optimising cold starts.

Around the platform. I also built a HIPAA-compliant audit-log service in a month, which unblocked the European release and grew revenue by 17%; an adapter between Catalyst and Zoho Cron for custom cron expressions that reached 35% adoption in three months; and error alerting by feature context that cut issue resolution time by 30–40%. As a project trainee I added environment variables for functions, which addressed 60% of user tickets, and Node.js 16 support with 12% faster cold starts.