Harsh K.

* Driving Meta’s company-wide Capacity & Efficiency program. Enabling engineering teams to deliver measurable reductions in MW consumption through targeted technical optimizations across the full stack to meet Meta’s capacity requirements for AI. * Building and deploying automated tooling to identify efficiency wins opportunities and detect regressions, integrating insights into engineering workflows for proactive mitigation. * Launching Efficiency Champions, a company-wide program with platform and product group leaders, to create a framework for efficiency and regression prevention and fostering a culture of compute efficiency to help Meta archive product goals amid datacenter capacity constraints.

* Spearheaded the onboarding, development, production hardening, and operationalization of database TiDB at LinkedIn as a 0→1 distributed database platform of choice, ensuring seamless integration into the company’s infrastructure. Showcased progress on the program by demo’ing features to the CTO and his staff on a bi-weekly basis. Drove discussions and influenced tradeoffs in design choices for scaling, federation, and others. Worked cross-organizationally to onboard internal systems, enabling the company to consolidate storage infra to the next gen offering. * Developed and delivered a 0→1 high leverage developer productivity platform Euler on TiDB that provided LinkedIn product engineers the agility to onboard entities and queries on Linkedin.com in a matter of hours as opposed to weeks and months saving engineering toil. * Initiated and led a program to migrate ~2000 Samza jobs (legacy stack) to Flink on Kubernetes enabling ML engineers to reduce toil by unifying streaming and batch jobs and utilize advanced features for real-time feature generation. Led efforts to synthesize technical solutions for different categories of jobs and drove cross-organizational alignment to prioritize the migration. Presented progress to the VP of Data Infra and other stakeholders. * Drove a large-scale, cross-organizational effort to migrate a set of infra systems (TiDB, Samza, Flink) to a new stack (provisioning, maintenance, decommission) on the LinkedIn on-prem cloud. Initiated and led a program to develop and drive widespread adoption of the high-performance scoring and ranking engine, Quasar, across verticals including Feed, Ads, and Recommendation Systems (e.g. People You May Know). Drove design considerations discussions for integration into each vertical.This eliminated custom scorers/inference engines (~40) per vertical that caused fragmentation. Quasar’s standardized DSL enabled leverage for reusability across verticals.

* Drove a company-wide security hardening program by scanning 100% of Intuit’s hosted cloud platform (~30K hosts) for 3rd-party software vulnerabilities, with results continuously monitored via a dedicated real-time dashboard. * Initiated a cross-company governance program for technology standards, driving a large-scale change management effort to align technologies with capability area standards.

EPM for a major Apple initiative in education (http://www.apple.com/education/preview/) EPM for a service called Device Enrollment Program (DEP) * Device Enrollment Program is an enterprise device management cloud solution. It was featured in WWDC in 2014 (1:06:37). * Managed a project that led to the expansion of DEP to 26 countries and 20 new languages. This project involved many cross functional teams and required planning and coordination. * Managed a project that expanded Device Enrollment Program to include reseller channels, in addition to Apple direct. * Led effort to gather marketing and business requirements, organize and kick off cross functional initiatives. Worked with cross functional engineering, marketing, legal, privacy, AppleCare and Human Interface teams to derive plans of record. Organized project plans and broke it down to components and held daily scrum meetings. * Identified risk factors and produced mitigation plans. * Worked with various teams to prioritize and re-prioritize features to fit the constraints of resources and time. * Communicated status to executives up till CEO's direct reports * Provided reports and gave roadmap presentations to cross functional teams and executives. * Lead an effort to on-board several customer systems on to a technology platform that can auto provision and de-provision instances based on load during peak seasons. This has led to operational cost savings for Apple.

Search & Media Sciences: • Search Sciences team working with Yahoo! Search. Owned the design, development and launch for MLR (machine learned ranking) models for relevance improvements • Worked with Yahoo! Search Engineering team to design, develop and deployment of searcher changes for improving relevance based on our ranking algorithm. • Developed the project and launch plans for MLR candidates. Presented the launch recommended deck to the executives. • Performed deep dive analysis to determine root cause of bad search results for queries. • Delivered the Full Business Insight/Revenue estimation pipeline for the Traffic Shaping project working with CORE/UDA team. • Optimized models for performance through early exits. Developed and generated the proxy dictionary for the US market. • Part of Global Relevance Team that won Yahoo Super Star Award for relevance improvements. Responsibilities included understanding business requirements, breaking down asks and prioritization, project planning and execution. * Trained the final ranking models for YST International Markets such as VN, IN, TW, etc * Cash Award for delivering the final ranking for Y! TW.

Applications on Demand Technologies: • Managed large scale internet application roll-outs including all project planning, cross functional coordination, resource and time allocation, risk mitigation, and execution during go-live. Design and architecture changes. • Managed several infrastructure maintenance projects. • Fully automated a manual and chaotic build/deployment process. Implemented daily and on-demand builds (at the click of a button). Web interface provided build metrics, logs, and list of changes. Published standards for source management including branching, tagging, and merging.

Computer Vision: Detecting Deformable Objects in motion using Least Squares Pictures and Trails: Computation of Shape and Motion From Perspective Image Sequences NSF GRANT IRI-9201751 On leave

Distributed Systems, fault tolerance, reliability, scalability Unreliable failure detectors: solving for Consensus in asynchronous systems with crash failures DARPA/NASA Ames Grant NAG-2-593.