Apr 2021 – Jan 2024 · 2 yrs 9 mos · San Francisco Bay Area · Hybrid

• I led the Instagram Growth ML team working on serving personalized account recommendations to users to help grow and retain the Instagram user base. I grew the team from ~6-7 to 15+ engineers. Our work included working on complex and interesting problems on Graph Learning, data mining, deep learning, causal inference, etc. Our services drove 10%+ retention and sessions in addition to 7%+ revenue. We have been able to achieve such a high impact by balancing short-term projects around sourcing, ranking, deep learning, etc., and long-term modeling using causal/correlation methods to understand the retention impact of our services.

Mar 2019 – Apr 2021 · 2 yrs 1 mo · San Francisco Bay Area · Hybrid

Jan 2018 – Feb 2019 · 1 yr 1 mo

May 2016 – Dec 2017 · 1 yr 7 mos

• Working in the small cell team at Ruckus.

Nov 2012 – May 2016 · 3 yrs 6 mos · Santa Clara

• Worked on research, algorithm & IP development and standardization of wireless communications,
• networking and connectivity technologies (5G, mmWave, IoT, LTE/LTE-A, WiFi, etc) as part of the Next Generation and Standards team at Intel. Projects included PHY and MAC layer design for device-to-device communications, mmWave, and Narrow-Band Internet of Things. I authored/co-authored 20+ invention disclosure filings, 10+ patent applications, 3 conference papers, 1 journal paper and 1 book chapter while at Intel.

May 2012 – Sep 2012 · 4 mos · Santa Clara, CA

• Designed and implemented an end-to-end system for adaptive bitrate video streaming over eMBMS, the broadcast delivery method in LTE-A, for optimization of video Quality of Experience (QoE).

Jun 2010 – Aug 2010 · 2 mos · Berlin Metropolitan Area

• Designed and implemented a coding-based solution to mitigate video disruption during switch-
• ing from unicast (e.g., cellular or WLAN) to broadcast wireless channels (e.g., MediaFLO, DVB-
• H) while streaming live video to mobile devices.

Aug 2007 – Sep 2012 · 5 yrs 1 mo · Notre Dame, IN

• We used tools from iterative optimization algorithms from statistics to solve coding problems in communications and control. Specifically, stochastic approximation are a class of iterative optimization algorithms used to compute the zeros or min/max of functions from noisy or imperfect observations of the function. By mapping communications and control problems to such functions, we designed algorithms for communicating across simple networks (for e.g., point-to-point, relay/cascade) with noiseless as well as noisy feedback, while also computing fundamental bounds on communication rates needed for stabilizing unstable control systems across such networks.
• MS thesis: Network Communications with Feedback via Stochastic Approximation (http://www3.nd.edu/~jnl/pubs/ukumar-ms-nd-2009.pdf)
• PhD dissertation: Feedback Coding Schemes for Control over Gaussian Networks (http://www3.nd.edu/~jnl/pubs/ukumar-phd-nd-2012.pdf)

May 2006 – Jul 2006 · 2 mos · Greater Edmonton Metropolitan Area

• Designed and implemented an algorithm to achieve time synchronization for a new preamble detection method proposed for an uncoordinated direct spread spectrum system using a modified
• packet format.