OnGrid Engineering Leadership

10xKafka throughput across 13 topics, no added infrastructure
99.95%uptime, with AWS costs down 22%
40% / 35%faster release velocity; fewer production incidents
13production database upgrades, zero unplanned downtime
8.5 hrsrecovered per engineer per week in a 15-engineer AI-assisted development pilot
4engineers promoted into senior roles

Challenge

Regulated BFSI clients, including banks, leave little room for failure. The job was to ship faster and have fewer incidents at the same time, get more out of the existing Kafka and AWS estate, and carry out production database upgrades without disrupting clients.

Approach

Kafka: fix the code, not the fleet

I diagnosed and personally fixed a JVM concurrency bottleneck in code, raising throughput 10x across 13 topics with no added infrastructure.

Delivery

I introduced weekly releases, stronger CI/CD controls and clear production ownership.

Database upgrades

Thirteen production upgrades went out through controlled MySQL and Debezium CDC migrations.

Cost and reliability

RDS rightsizing, CloudWatch observability and FinOps.

AI-assisted development, measured

I designed and measured a pilot with 15 engineers before any wider rollout. The measured result led to org-wide adoption.

People and compliance

I ran appraisals for 10 direct reports, including development plans and underperformance cases, and led the SOC 2 Type 1 implementation.

Results

Increased Kafka throughput 10x across 13 topics with no added infrastructure.
Improved release velocity 40% and cut production incidents 35%.
Cut AWS costs 22% and raised uptime to 99.95%.
Delivered 13 production database upgrades with zero unplanned downtime.
Recovered 8.5 hours per engineer per week in the pilot, leading to org-wide adoption.
Promoted 4 engineers into senior roles.
Led SOC 2 Type 1 implementation.

Stack

Java
Spring Boot
Kafka
Debezium CDC
MySQL
AWS RDS
AWS CloudWatch