YugabyteDB is an open-source, high-performance distributed database with PostgreSQL compatibility, designed to deliver transactional consistency and resilience for mission-critical applications at global scale.

I work in the Language and Relational Technologies (LRT) team that is responsible for query planning, execution and optimization. I drive write-path development in the query layer, bringing the rich feature set and robust transactional semantics of Postgres to the world of distributed SQL. My focus areas include:

⇢ Performance - optimizing query execution for write workloads (inserts, updates, upserts, deletes, triggers, constraints) by tweaking Postgres' traditional single-tuple volcano approach.

⇢ Demonstrable correctness - ensuring that the optimized execution adheres to Postgres' consistency and isolation guarantees, and by extension, the broader SQL standard.

Given the inherent complexity of executing queries in a distributed environment, I’m also passionate about advancing query explainability and observability to make these systems more transparent and diagnosable.

Part of the CMU-DB Research Group, advised by Prof. Andy Pavlo.

Department: Nimble (acquired by HPE)

I worked on the Log Structured File System (LFS) and the Garbage Collection services of NimOS, Nimble’s operating system based on their patented CASL architecture.

The LFS packs variable-length data blocks (from data writes) into logical segments, each of which is persisted in physically contiguous units of storage. The LFS is additionally responsible for data compression, encryption and integrity.

As part of the internship, I developed a distributed simulator for NimOS’ Segment Layer, a service that provides a consistent virtualized interface for the underlying hybrid hardware storage. This was part of a larger internal rewrite of the stack to support cluster aware space (storage) management. In particular, this project decoupled the development lifecycles of the LFS and the Segment Layers.

Teams: CALM (Cloud Application Lifecycle Management), Prism (Management Plane Infrastructure).

Part of the core development team of Calm SaaS (link below), a fully-managed, subscription-based application management offering, designed to expand Calm’s TAM to customers who do not own a Nutanix appliance.

Primarily responsible for the design and development of the SaaS platform on public cloud including:

⇢ Networking stack: Secure flow of traffic from public cloud to private VPCs and data centers, and vice-versa; firewalling and proxying. Also worked on network security, and SSO integration with the Nutanix portal.

⇢ On-premise agent: A smart agent that manages connectivity to network blackboxes (ie. VPCs, data-centers), with self-healing and self-upgrade capabilities.

⇢ Upgrades: An ecosystem to facilitate the regular roll-out of transparent upgrades to the Calm platform as well as to deliver critical security patches and functionality improvements to the On-premise Agent.

⇢ Monitoring and Telemetry stack for the platform, including real-time actionable alerts for a variety of scenarios including resources crunches, services downtimes and semantic failures in the product/platform.

⇢ Builds and deployments: Build pipelines for the platform components, and automated provisioning and lifecycling of Calm instances on-demand.

Worked on a variety of projects/domains across the Nutanix stack. Highlights:

⇢ Support for Same-site replication and recovery of VMs for disaster recovery/business continuity.

⇢ Segmentation and isolation of control plane network traffic from user/management traffic.

⇢ An automation framework for specifying and generating Calm blueprints on the fly.

⇢ Performance benchmarking of Nutanix API infrastructure.

Conducted 10+ bootcamp and technical deep-dive sessions for new hires, on 'Application Lifecycling and Governance' in hybrid cloud.