-
Adobe Story
(December 2015 - January 2017)
- Adobe Story is a web based editor for creating screenplays and script writing, integrated with scheduling and reporting for production process.
- Designed and built the architecture for a real-time collaborative script editor on browser, based on web-sockets and NodeJS.
- Implemented several data models associated with script such as camera shots and recording tapes. Implemented pagination and undo-redo on the document that works with several authors editing in real-time.
- Built an optimised overlay renderer that renders any UX element on top of the script annotated to some script text.
-
Adobe Premiere Clip - Android and Web Application
(October 2014 - November 2015)
- Adobe Premiere Clip is a video editing application with 100k+ installs.
- Created MOV file playback support on Android which does not natively support it. For this I built an efficient video/audio sample loader that interacts with decoders in other threads to give a seamless video/audio playback. Built an efficient Thumbnail extractor that can fetch frames for Ultra HD videos.
- Designed and built backend and UX for a video player and an editor that can edit a sequence of videos and images. Built features such as trimming, transition between clips, sequence editing, lighting filters such as exposure, highlights, shadows, audio controls etc.
- Designed internal data models for Projects and Sequence of media files. Created framework for interaction between UI, playback and rendering threads on Android.
- Built an efficient Rendering pipeline that applies filters to frames on the fly during video playback. Built a single shader in OpenGL that implements several filters by loading LUT (Lookup Table) data in an intelligent way.
-
Adobe - Media Grouping Algorithm
(July 2014 - September 2014)
- Intention of the algorithm was to group a user's media files in a way such that those groups represent memories and events in user's life. Each memory or event can have further sub-memories and sub-events.
- The algorithm analyzes media metadata and groups the files recursively for navigation. The algorithm uses time and optionally location (prediction is done if unavailable) to build a tree of user's media files. Built data structures to efficiently fetch information of a node representing an event.
- Built client-server interface to fetch the groups and designed a database to hold result data.
-
Adobe ATL - Customer Lifetime Value
(May 2013 - July 2013)
- At Adobe ATL (Advanced Technology Labs), I identified the current research landscape on the business metric called Customer Lifetime Value, came up with and solved a challenging problem.
- Current mathematical models for CLV use several customer parameters to predict their future profitability. With the growth of user's activity on social media websites, it becomes interesting to see how their activity on social networks can be leveraged to improve these models.
- My work revolved around predictive models and machine learning techniques to improve upon these models and evaluate the profitability of Adobe customers on Behance (social network for creatives).
-
Prediction of Product Sales and Popularity Using Social Media Data
(August 2013 - April 2014)
- The project aims to develop a predictive model that analyzes the relationship between user ratings of the products and user characteristics and perform product recommendations.
- The project was done as part of my undergraduate thesis.
-
Application of Kalman Filter to Smoothen GPS Navigational Data
(September 2013 - October 2013)
- Studied in detail, the mathematics and intuition behind the workings of Kalman Filter.
- Implemented a filter that inputs GPS navigation data obtained from a ballistic missile containing white gaussian noise and outputs an estimate of the actual smoothened path taken by the missile.
- Term Paper on the project - Application of Kalman Filter on GPS Data.pdf
☰
