Oct 2022 – Dec 2024 · 2 yrs 2 mos · Hybrid

• Proposed a DataQuality framework and scoring which helps identify data freshness issues, inconsistencies across datasets, semantic errors. Taken-up by the DQ platform team to be integrated into their existing framework.
• Designed an Orchestration platform which orchestrates multiple pipelines execution for multi-env staging, IA, backfills, delivered with high-quality and E2E completion
• Designed and led team to deliver events-ranking solution which serves 3 different business usecases, and reduces E2E Adtech latency by 37% (typical) - 50% (worst-case).
• Designed and led team on a pioneering new approach where AS can completely own and develop the Measurement KPI stack (multi-semester effort). KPI system is business-critical (guides annual Paid Marketing Spend of Uber), complex system (computes 15+ RoI metrics across 20+ dimensions, has 15 upstream data sources). The new system is a 5-layered architecture with ~ 15 components and enables AS to make changes in a self-serve manner.
• Led team to build and deliver multi-env system which allows the DS stakeholders to run upto 8 experiments in parallel in staging env.
• Reduced oncall noise from 120 to 35 p.w, oncall load by more than 60%.
• Fixed various bottlenecks and data-skew handling for a complex batch-processing system which processes 5.87 B datapoints spread over 1.2 B categories.
• Advocated and led team to productionize critical tech-debt reductions which were stalled earlier, which improved developer velocity by 50%

Oct 2021 – Sep 2022 · 11 mos · Hybrid

• Led the Data Lifecycle Management (DLM) Charter completely out of India (TL & EM of the DLM team).
• Defined the vision, designed a Unified Platform Service for Data Lifecycle Management to perform various data lifecycle operations such as time-based deletion, backup, tiering, archival, user account deletion, and developed it with the team.
• The DLM systems handle policy-based deletion of 35 PB data (logical) each month.
• The same platform & design is still in use after 2 years, and extended to more usecase scenarios.
• Primary author of a published Uber Engr Blog on DLM.

Aug 2021 – Oct 2021 · 2 mos · On-site

• Guided team to deliver the Backend Consolidation of two Azure Services (Azure Import/Export, Databox).
• Designed and delivered on the Azure Import/Export passthrough mode, which gives a seamless user experience with the existing APIs while leveraging the Converged stack in the backend.

Dec 2015 – Aug 2021 · 5 yrs 8 mos · On-site

• Discovered and fixed critical gaps in various products – security authentication, non-billing of offering, secret-rotation handling
• Implemented a host of product and process improvements in ImportExport service which drastically improved the release effort from ~ 2 weeks to 2 days
• Led team in convergence effort to serve different offerings from a single service, added support for multiple sub-workflows, designed data passing mechanism in the workflow
• Single-handedly implemented and delivered Secrets autorotation capability in ImportExport service, and Azure Stack Edge service
• Orchestrated, led and implemented the GDPR compliance effort in team, was the Subject-Matter-Expert (SME) on GDPR in the team
• Implemented resource handler component to provide a generic resource representation which can be used by ARM RPs, reused in all the Databox family of products

Mar 2015 – Nov 2015 · 8 mos · On-site

• Developed support for locally-pinned volumes for StorSimple product

Jan 2013 – Feb 2015 · 2 yrs 1 mo · On-site

• Core member of Health & Fitness vertical in the redesigned MSN (www.msn.com)
• Implemented critical features, created server-side web-services using ASP.NET (C#) and client-side development (HTML5 Canvas, JQuery / JavaScript).
• Delivered two complex features of Health vertical (Symptom Checker, 3D Body Maps)
• Worked on various parts of Health vertical and common page types
• Performance specialist for ensuring enhanced user experience (reducing page load time, improving RPS, memory optimization, etc)
• Designed module-view-caching to improve Server RPS from 15 to 52 RPS
• Reduced image sizes by more than 50% (low-res & high-res)
• Improved zoom performance by 3.5x
• Proactively worked with team members to improve performance of their features
• Came up with an innovative image compression scheme to represent area maps (~25% size reduction compared to Png, visibly faster performance for region highlight)
• Developed few important features of Win 8 Health & Fitness App

Jan 2010 – Dec 2012 · 2 yrs 11 mos

• Developed multi-screen transcoder for US-based startup, 3G streaming server / client for Spanish customer, handheld media player on RTOS for Japanese customer
• Designed and developed base-child node architecture, which resulted in 50% code-size reduction (nodes are similar to Filters in DirectShow)

Jan 2007 – Jan 2010 · 3 yrs

• Architected, designed and developed Multimedia System Framework. This is the embedded-system equivalent of DirectShow on Windows, and GStreamer on Linux.
• Highlights: Optimal memory utilization, Interface to specify & control the thread architecture, Supports variety of use-cases (Player / Recorder, Streaming Server / Client, Transcoder, Multi-screen)
• Developed a unit-test framework for unit testing multi-instance and multi-threaded systems.
• Designed and developed a full suite of Container format Parser and Creator components.
• Achieved goal of zero-maintenance of these components for a variety of use-cases through extensible design, thorough unit testing (code coverage ~ 90%), complete documentation (Req. Spec, User Manual, Design Doc, Test Plan / Test Report).

Jan 2005 – Jan 2007 · 2 yrs

• Designed and developed Multimedia Cache module, which implemented reactive-proactive caching. This was a key enabler for the Media Player to play network content (for connected home entertainment systems)