Jul 2023 – May 2025 · 1 yr 10 mos · United States

• I am the Engineering leader for Coupang Play - premium video content streaming service from Coupang (https://www.coupang.jobs/en/teams/coupang-play/). Coupang Play is the second largest (after Netflix) video streaming service in Korea with 8M+ MAU (https://www.wiseapp.co.kr/insight/detail/528). Play experience is available on Android, iOS, Web and smart TV platforms - service offerings include on-demand content (licensed + originally produced), transactional content (rent or purchase), live sports and news as well as community experiences such as group watching.
• My job responsibilities include product engineering (e.g., user experience, personalization, payments, video playback, identity and live events), backend and platform engineering, machine learning, video streaming, as well as content ingestion and management systems.
• We are a fast growing team of nearly 100 developers dedicated to the goal of building world-class video entertainment experiences making Coupang Play the definitive destination for digital content.

Jun 2022 – Jul 2023 · 1 yr 1 mo · San Jose, California, United States

• Our team supports the AWS Cloud Infrastructure and the media processing platform for Frame.io
• # Founded (worked with customers and stakeholders to define problem statement), led development, and productized Frame.io Storage Connect solution (https://blog.frame.io/2023/09/13/introducing-frame-io-storage-connect/), starting from zero
• # Tactical projects and strategic initiatives to improve availability of Frame.io service. Impact includes decrease in service outages from hundreds of minutes of downtime in first half of year 2022 to nearly zero downtime in the second half of the year
• # Identified and implemented cost improvements decreasing AWS costs by one-thirds. These savings helped increase overall gross margin of Frame.io business by more than 7 percentage points

Oct 2019 – Jun 2022 · 2 yrs 8 mos

• Our team runs the largest live video streaming origination platform in the world deployed via globally available services on AWS as well as bespoke metal infrastructure. We develop systems to acquire live video streams from the community of Twitch broadcasters, perform real-time processing (e.g., video trans-coding), and make them available for global distribution at scale. A related tech blog article could be found at https://blog.twitch.tv/en/2022/04/26/ingesting-live-video-streams-at-global-scale/
• # Supported an organization of about 40 (grew team by 2X) team members - hired and promoted key
• individual contributors and people leaders
• # Played key role in productization of Amazon Interactive Video Service (IVS) offered by AWS in July
• 2020, followed by ongoing rapid feature release iterations
• # Led dedicated investment to improving reliability, scalability and maintainability of the Twitch video platform services (5X decrease in number of system outages year over year)

Oct 2014 – Oct 2019 · 5 yrs · San Francisco Bay Area

• # Technical lead for the cloud-based content ingestion system developed for the SMPTE Interoperable Master format (IMF). Recurring savings resulting from implementation of IMF estimated to be hundreds of millions of dollars
• # Lead contributor to the IMF standardization effort at SMPTE forum. Lead developer for Netflix OSS project “Photon” (https://github.com/Netflix/photon). Photon is used by 100% of IMF authoring tools worldwide. Recipient of the 2019 HPA Engineering Excellence Award for OSS project Photon
• # Founder of the "Media Intelligence" initiative at Netflix - leading a team of computer vision and natural language processing experts skilled in using deep learning methods for various media understanding problems. Examples include object/text detection in images and video, video shot change detection, and automated artwork generation. Models developed by the team power the netflix.com user experience
• # Supervised the development of “Netflix Media Database” - a multi-tenant, highly available, schema-on-write, and a highly queryable data system for technical metadata associated with multimedia assets at Netflix. Implementation of the data system uses both Netflix internal as well open source components including Apache Cassandra and Elasticsearch. This data system powers all of the video encoding workflows at Netflix
• # Participant in W3C Timed Text Working Group (TTWG). Contributed syntactical and rendering primitives necessary for supporting Japanese subtitles. Recipient of the 67th Annual Technology and Engineering award (Tech Emmy) titled "Standardization and Pioneering Development of Non-Live Broadcast Captioning" (January 2016)
• # Team lead for implementation of a i18n grade (Unicode based) timed text (subtitles) ingestion
• and trans-coding system. This was developed for the world-wide Netflix launch in January 2016
• # Implemented ISO MP4/DASH based video packaging solutions for (a) DolbyVision HDR (HEVC); (b) and VP9 bitstreams

Jan 2013 – Oct 2014 · 1 yr 9 mos · San Francisco Bay Area

• # Project lead for implementation and deployment of a cloud-scale Linux based workflow for ingestion of Apple iTunes specification compliant media files (Apple ProRes video + PCM audio) replacing the legacy closed source Windows based workflow. Workload migration from legacy to new system was done with zero downtime and zero regression
• # Lead contributor for the overhaul of the Netflix media trans-coding pipeline transitioning from a data center to an AWS based implementation. Co-ordinated fixing of all audio and video processing issues and ensured successful transition to next-gen video encoding platform
• # Implemented ISO MP4/DASH based video packaging solution for client-side trickplay. This solution powers the Netflix playback experience on iOS platform
• # Implementation of video encoder and decoder ffmpeg plugins
• # Replaced of end-of-life Windows DirectShow media processing filter components with those from a new vendor

Oct 2011 – Jan 2013 · 1 yr 3 mos

• # Responsible for cloud-deployed (AWS) real-time audio and video mixing and processing services for the TenHands collaboration product. Contributions include:
• (a) real-time bandwidth adaptation for media streams including integration of a novel delay-based congestion control algorithm
• (b) jitter buffer management
• (c) audio/video quality improvements
• (d) adaptive lip-sync algorithm
• (e) enhancements to the base RTCP stack implementation and
• (f) integration of audio codecs such as Opus
• # Contribution of accepted patches to open source webRTC project
• # Development of automated test-suite for TenHands web application using Ruby-based Watir framework
• # Managed audio and video (VP8) compression and transport components for the web-browser based TenHands real-time client application
• # Added (a) security (using SChannel/openSSL) to websockets implementation and (b) proxy traversal capability to the browser based TenHands client application

May 2010 – Oct 2011 · 1 yr 5 mos · San Francisco Bay Area

• # Advisor for the video-encoding, image processing and video streaming component of the solution

Mar 2010 – Oct 2011 · 1 yr 7 mos · San Francisco Bay Area

• # Development and maintenance and of (proprietary) real-time multimedia framework developed for TI platform (C64x+ core) using TI DSP/BIOS primitives
• # Development and maintenance of unit test framework for testing of product software for TI platform
• # Architectural and feasibility analysis for implementation of real-time video and speech analytics algorithms on various hardware platforms including TI and Intel
• # Development of a high-definition H.264 SVC video codec on an hybrid (DSP + FPGA) platform for a video-conferencing/tele-presence application

Apr 2008 – Mar 2010 · 1 yr 11 mos · San Francisco Bay Area

• # Development of advertisement operations (adOps) UI implemented in Ruby on Rails (RoR) framework. The UI enables configuration and control of advertisement campaigns including management for advertisement agencies, advertisers, banner, video as well as HTML creatives
• # End-to-end management of server release (November 2009) including engineering development co-ordination, build management for engineering testing and QA, and live upgrade of production system
• # Development, maintenance and performance tuning of a Content Ingestion system implemented in J2EE framework. This system is used for ingestion and trans-coding of video, image, and text (blog) stories provided by third-party content providers in MRSS format or through an FTP dropbox
• # Development and maintenance of a Content Management System (CMS) UI plus associated MRSS feed (implemented in Ruby on Rails (RoR) framework). The UI enables content editing/aggregation for ingested content as well as a drag-and-drop framework for the purpose of building a “program guide” with channels containing aggregated content
• # Performance optimization of program guide (an XML document) delivery system including code optimizations for improving loading times for client smart phone handset devices and server side caching (Apache caching) to reduce load on delivery system
• # Upgrade of Ruby on Rails (RoR) based web server from Rails 1.2.3 to Rails 2.1.1

Jan 2007 – Apr 2008 · 1 yr 3 mos · San Francisco Bay Area

• # Part of next-generation graphics processing unit (GPU) pipeline development effort. Tasks included development of test applications (using an in-house openGL/Direct3D like API and Cg programming language) for functionality and performance testing of different GPU pipeline stages
• # Part of a low-power multi-format video decoder (H.264, VC1, MPEG-4, and JPEG) and AES encryption/decryption hardware pipeline development effort. Tasks included:
• (1) decoder pipeline architecture development
• (2) architecture performance analysis
• (3) hardware engine modeling, and
• (4) test suite/regression scripts for hardware module verification

Aug 2005 – Jan 2007 · 1 yr 5 mos · San Francisco Bay Area

• # Part of a low-power VC-1 (WMV9) video decoder development effort on an ARM core + hardware acceleration platform. Tasks included:
• (1) development of overall decoder pipeline
• (2) architecture for constituent hardware modules
• (3) hardware engine modeling (using SystemC C++ library)
• (4) test suite for hardware module verification
• (5) Responsible for decoder compliance testing and
• (6) overall error handling architecture for decoder
• # Part of a low-power H.264 video decoder development effort on an ARM core + hardware acceleration platform. Worked on different aspects of decoder including:
• (1) hardware-software partitioning
• (2) software architecture and performance optimization on ARM7 platform
• (3) hardware engine architecture
• (4) test suite for hardware module verification and
• (5) decoder compliance testing
• # Proficiency in multimedia file formats including MP4, MOV, 3GP, AVI and ASF
• # Technical evaluation of 3rd party audio/video codec IP on ARM platform