Working to strengthen and scale the integrity systems that underpin all of Meta's platforms.

I joined the organisation when the product and the execution needed to scale rapidly. I started an initiative to improve engineering velocity while reducing the number of customer-impacting errors by redesigning safety mechanisms, developer tools, as well as engineering methods. I then crafted the vision and started a multi-year scaling initiative to enable some of AWS’ largest customers to test their business critical services spanning tens of thousands of AWS resources across Availability Zones and regions. I also built the foundational systems to support the first observability capability offered to customers. This serves as the basis for several potential product features in the area of observability.

Principal Engineer at Amazon Robotics AI

I joined the organisation to help mature and scale the software systems across the edge and the cloud for the order consolidation robotic workcell. The following are a couple of the highlights from this tenure:

* Led the edge-computing team’s transition out of the prototyping phase to develop the crucial building-block of microservice orchestration at the edge, which was a complex, irreversible design decision for the organisation.

* Guided the redesign of the ML feature-computation subsystem to reduce item classification failures from nearly tens of thousands monthly to nearly zero, with the latency reduced from tens of seconds to milliseconds at the 99th percentile.

Principal Engineer at Amazon Go (JWO)

I was the founding engineer of a new org-wide function to redesign the infrastructure to scale the store footprint from the initial "Just-Walk-Out" computer-vision enabled store, to tens of such stores over the next few years. As part of that, I led the following initiatives:

1. Re-engineering of the infrastructure to enable true elasticity to support variable traffic, and on-demand provisioning globally. This entailed redesigning the software building blocks and fundamental assumptions of fault-tolerances across tens of micro-services. I then helped scale the metadata system, infrastructure automation, and best practices to further accelerate the launch of the stores rapidly across geographies.

2. Redesign foundational software used across JWO's micro-services to support fully automated activation, and touch-free end-to-end computer vision simulation to prove correctness automatically. It enabled decoupled vision software activation across regions and teams, eliminating the need for cross-service coordination.

3. Redesign of the in-store software to support fully automated activation and vision-hardware placement optimisation.

The above initiatives resulted in true infrastructure elasticity, fault isolation, and operational excellence for new store while reducing hundreds of hours of engineering effort and thousands of dollars in operational expense.

As one of the founding engineers, I built InMobi’s first scalable, high throughput, low latency ad-serving system that continues to serve billions of requests daily. As the system scaled to support multiple businesses lines and engineering teams over the years, I joined the CTO’s office where I was the advisor to the CTO in strategic decisions on technology investments, organization structure, leadership hiring and nurturing, and technology due-diligence for M&As. Over nearly a decade, I also transformed the engineering methodology to help teams to operate with greater agility and autonomy to support new products. Key achievements include:

1. Design and implementation of InMobi's next generation, cloud native, privacy aware, terabyte scale multi-tenant user data management system.

2. Transformation of the engineering culture to achieve operational excellence and agility. Introduced SLO definitions, metrics-centric engineering, a move to infrastructure agnostic service design. Also entailed building deeply integrated services for container deployment powering thousands of containers, and services for metrics collection & monitoring. The erstwhile operations function was revamped into a lean SRE function that owned the above services globally.

3. Coauthored the company's cloud transformation strategy, implementing hybrid infrastructure that enhanced engineering agility and cost efficiency

4. Conceptualized an ML model management and serving system similar to Uber's Michelangelo, built on Apache Spark. This led to better infrastructure utilisation and efficiency, maintainable data pipelines, 30% lower turnaround time processing terabytes of data, and reduced training time from hours to minutes.

5. Introduced a multi data center data assimilation system that was resilient to failures, and was capable of assimilating data updates of varying granularity and frequency.

6. Devised a system to join monetisation event streams of varying scale and data volumes.