Oct 2025 – Present · 7 mos

• 1. MQTT platform initiative planning for every half.
• 2. Pair with teammates to unblock them & ensure timely delivery of our planned initiatives.
• 4. Being the first on-call responder, help close incidents involving our MQTT platform.
• 5. Own my platform's business metrics & ensure its availability.
• 6. Undertake individual initiatives in every half - H2 2025 - Move HTTP polling based driver location & ETA tracking on consumer app to MQTT platform.
• 7. Conducted 10+ in-person on-campus interviews for IIIT Hyderabad & BITS Pilani.

Apr 2024 – Sep 2025 · 1 yr 5 mos

• 1. Led & executed the cloud migration of GoTo group's MQTT infrastructure from GCP to Tencent Cloud as part of the company-wide initiative to move its complete infrastructure to Tencent & Alibaba Cloud, thus contributing to the overall group-wide cloud costs savings of more than 55%.
• 2. Saved $25000 USD in annual compute costs on GCP by autoscaling VerneMQ using HAProxy-Consul integration.
• 3. Reduced CPU usage by 33% on the driver MQTT broker cluster and 40% on our events service by implementing a topic-based sampling feature on VerneMQ.
• 4. Migrated all 5 MQTT infra supporting Golang services from self-managed K8s to a centrally-managed K8s cluster, reducing operational overhead.
• 5. Mentored & onboarded 3 new joinees in my team, leading to increased task delivery and stronger ownership of responsibilities.
• 6. Collaborated with the new hires to enhance stack observability with Prometheus metrics and standardized logging using Barito, enabling faster debugging and comprehensive visibility of all onboarded use cases.
• 7. Served as the primary on-call responder for all Courier product incidents and support requests.
• 8. Conducted 28+ online interviews for L1 - L4 roles including bar-raiser & machine coding rounds.
• 9. Participated twice as a coach in GoTo’s Engineering Bootcamp at Gojek HQ, Jakarta, Indonesia.
• 10. Evangelised about Courier & Gojek open source at Mentor’s summit, Google's office in Sunnyvale, California.

Oct 2022 – Mar 2024 · 1 yr 5 mos

• 1. Enhanced durability & availability from 99.9% to >99.99% of VerneMQ brokers by integrating Redis as an offline store, preventing message loss and eliminating the need to delete the complete state when cluster node goes down.
• 2. Traffic remains unaffected and migrates affected nodes’ connections to others in the cluster without any state loss.
• 3. Scaled broker for 3M+ rpm messages to onboard driver pings use-case by optimising shared subscription routing.
• 4. Added support for QoS1 without persistence/retry, involving protocol changes on VerneMQ.
• 5. Migrated the entire MQTT infrastructure from a legacy Google Cloud Platform project to a new team-level GCP service project using Terraform, streamlining infrastructure management.

Sep 2021 – Sep 2022 · 1 yr

• 1. Owned deployment, scaling, and maintenance of VerneMQ broker clusters in production.
• 2. Implemented automatic cluster remediation, increasing availability to over 99.99% and reducing recovery time from 30 minutes to 4 minutes, without requiring manual oversight.
• 3. Added support for MQTT v3.1.0 sticky-sessions in HAProxy.
• 4. Collaborated on migrating the events plugin from HTTP/1.1 to gRPC, preventing broker choking and message loss beyond 3.2k messages per second.

Jun 2021 – Aug 2021 · 2 mos

• 1. Setup MQTT loadtest bench that generated 1M+ concurrent connections with 300K rpm messages - Medium for traffic patterns of use-cases like push notifications, driver mobile pings, order updates, chat, etc.
• 2. Scaled & tuned MQTT broker system(GLB, HAProxy, VerneMQ & WebHook Service) to handle the generated load.
• 3. Identified & resolved scaling issues in VerneMQ broker - 35% message loss beyond 3.2k message/sec, VM crash beyond 28K concurrent connections, etc.
• 4. Reduced developer’s effort to run a loadtest scenario and its run-time from 1.5 hours to 20 mins.
• 5. Saved the operating cost for load testing by 50%.