Constantin Dumitrascu

I managed a team of engineers while providing technical leadership to 30 people across multiple organizations and roles (security, networking, datacenter, logistics, project manager, product manager, and compliances).

As a tech lead and manager, I created Google's next-generation product, Transfer Appliance, a high-capacity storage device that enables you to transfer and securely ship your data to a Google upload facility. I worked on it from scratch and developed versions TA300 and TA40, scaling them to 300TiB for each server.

This is a hybrid project, and I owned the software and end-to-end experience, including the hardware, operating system specifications, data center operations, logistics, networking, and security. Transactions run long across multiple locations, both offline and on-premise. This was a fantastic engineering feat, and we overachieved on its three pillars – security, performance, and ease of use.

A project that empowers people to explore and discover a research field independently by identifying the most influential articles to read.
It supports 20 computer science fields—AI, crypto, networking, algorithms, and distributed systems, to name just a few.

I contributed everything—from identifying the need to sourcing the data, designing the systems, building it, launching it, and acquiring users.

I led the development and timely launch of a hybrid processing platform, both traditional computing and quantum, for long-running processing jobs. My work spanned from shaping the public APIs to back-end capacity planning with thousands of CPUs, integrating with SageMaker, guiding public documentation, setting up observability, leading five engineers, and coordinating another 20 to make estimates and deliver on time.

I designed, built, and deployed the Amazon retail monitoring system, a multiplexer engine ingesting and analyzing millions of events to detect problems with the retail website experience. This saves Amazon hundreds of millions of dollars across the globe, as it's the critical engine that keeps the site up. I provided technical leadership to a team of seven engineers while coordinating with multiple organizations, specifically AI and production engineers.
My most critical contribution is designing a back-testable system, reducing the test time by 1000 times, from months to minutes.

I led a team of six engineers while coordinating with eight XFN partners to redesign and rewrite Facebook's video encoding and delivery infrastructure.
While servicing 90 billion videos per day, I completely rewrote the delivery stack (20+ thousand lines of code) to improve the user experience without causing any outages. I've made these systems self-healing and added graceful degradation to withstand annual demand spikes without outage while reducing the encoder's capacity needs by 40% or 10K servers.

I invented, designed, and delivered a solution to a problem previously thought impossible: automated prediction and management of Facebook's heterogeneous fleets of hundreds of thousands of machines across ten data centers, and made back-end services resilient to disasters with a loss of four data centers, without over-provisioning them.
I led a team of five engineers and acquired my own internal customers while coordinating with the capacity-engineering organization and gaining their trust for such automation.

I built several parts of the portal, including file uploader, reporting, CRM, and mass email notifications. I also rewrote the automated machine-image scanning for vulnerabilities to validate that the AWS marketplace buyers purchase only safe products.