Feb 2021 – Mar 2022 · 1 yr 1 mo · Vancouver, British Columbia, Canada

• Responsible for delivering state-of-the-art personalization platform and player matchmaker for mobile e-sports

Jan 2018 – Feb 2021 · 3 yrs 1 mo

• Responsible for conceptualization, prioritization, and delivery of Bixby 2.0, Samsung's latest conversational AI Assistant. Working with a cross-functional team of UX, engineering, marketing, security and legal experts on multiple features in Bixby used by millions around the globe.

Mar 2017 – Jan 2018 · 10 mos

• Samsung Pay - Fintech & Mobile Payments
• Successfully launched Samsung Pay in Sweden in April 2017
• At the time of launch supported 3 out of top 5 banks(Nordea, Handelsbanken, Edenred) covering more than 60% of all payment cards
• Lead the technical efforts to onboard Swedbank, one of the biggest banks in Sweden, on Samsung Pay
• Led technical efforts to provide various other features - NFC card enrollment, App2App card enrollment with bank apps, SMS autofill for OTP.
• Led efforts to support multiple in-app promotions to increase usage of Samsung Pay
• Collaborated with the Marketing team in order to drive monthly active users and daily transactions

Sep 2015 – Mar 2017 · 1 yr 6 mos

• Deployed data center of Samsung Cloud service in 3 regions-US, SG, and AMS-together with a team of 8infrastructure engineers. Samsung Cloud serves hundreds of millions of users of Samsung devices.
• Data center Architecture- Participated in designing and deploying 3-tier network topology for Samsung cloud platform along with Load Balancers and WAF (Web Application Firewall).
• Mediator- Developed Mediator, a framework of Apache Mesos used for resource management and scheduling across the entire data center. Mediator supports resource scheduling with dynamic reservation and quota allocation. Co-developed Mediator Drill, an application to do integrated testing of Mediator.
• Hermes DNS Service- Co-developed Hermes, a highly available cloud Domain Name System service. Hermes effectively maps end-users to infrastructure running in Samsung’s cloud platform. Hermes also supports a visual editor where you can easily manage how end-users are routed to application endpoints.

May 2015 – Sep 2015 · 4 mos · Mumbai Area, India

• Computer Vision Lead
• Related Products- Developed ML model to retrieve similar styled apparels for product pages from thee-commerce inventory of 1 million apparels.
• Apparel Classifier- Implemented modified Knn-classifier to classify a dress based on Neck type (9 categories),Sleeve type (4 categories), Pattern type (3 categories) achieving 85.7% accuracy

Jan 2015 – Apr 2015 · 3 mos · Mumbai Area, India

• RemitMas aims to build a digital money transfer service that doubles as a savings tool for Latino immigrants in the United States and their families back home.
• Designed and Developed Web interface for the online transaction service.
• Won Dean’s Public Policy Challenge Grant competition of School of International and Public Affairs, Columbia University

Aug 2014 – Mar 2015 · 7 mos · Mumbai Area, India

• Worked on video streaming to Android devices in a classroom using wireless multicast and developed efficient and reliable techniques to stream videos in this environment.
• Conducted various experiments varying the wireless data rates, RSSI and different modes to measure throughput and loss rates, to analyse the feasibility of video streaming using WiFi.
• Developed a prototype Android application to send a file reliably from a server to multiple clients.

May 2014 – Jul 2014 · 2 mos · Suwon, Gyeonggi-do, Korea

• Researcher and Developer in Software R&D Center
• Proposed four Optimizations for Hadoop’s Distributed File System. Analysed and modified the source code of HDFS to find the performance bottlenecks and improve it.
• Replicator : Independent module to modify replication factor of a file in a cluster based on its access frequency.
• Disk Access Optimization : Achieved 6.1% improvement by analyzing the bottleneck in read operations at a datanode and implementing memory-buffered read to reduce disk seeks.
• Balanced Block Placement Policy : Modified Hadoop’s block placement policy to create a balanced cluster by accounting memory usage of nodes and racks while choosing a node to place a replica.
• Block-level Caching and Pre-fetching : Achieved 5.7% improvement in read operation by extending
• HDFS’s caching mechanism to a finer block-level caching and implementing pre-fetching of blocks.

May 2013 – Jul 2013 · 2 mos · Germany

• Defined a resource-aware semantics of concurrent programs that makes explicit use of cluster nodes, memory, and compute as well as data transfer times.
• Performed statistical analysis to determine a partial-order from the resource-aware semantics that optimizes the scheduling goal using solutions of mixed-integer linear programming problem.
• Drafted a research paper describing analysis and implementation of solution to scheduling problem under supervision of guide.