Aug 2022 – Feb 2025 · 2 yrs 6 mos · Seattle, Washington, United States

• » Worked across orgs in an advisory capacity to influence solutions and drive adoption of models.
• » Improved performance of agentic behavior in LLMs.
• » Finetuned large language models (LLMs) for Amazon AGI (upto 100B params).
• » Researched on smoother convergence of models during SFT and Model Merging strategies.
• » Worked on multiturn preference optimization of LLMs, a key problem in AI agents involving planning and execution.

Oct 2020 – Aug 2022 · 1 yr 10 mos · Seattle, Washington, United States

Mar 2017 – Jul 2019 · 2 yrs 4 mos · Greater Seattle Area

• I was involved in launching Alexa SmartHome v3 APIs in 2017 where I mainly worked on next generation OAuth services - https://tools.ietf.org/id/draft-hardt-oauth-mutual-02.html . This required working with four different organizations within Amazon and making changes to multiple services in a manner that ensures high security standards.
• 2018 was a highly productive year. In Q1 2018 I worked on improving the performance of a service that led to smoother service executions and eventually reducing infrastructure costs significantly. In Q2 2018, I was involved in building systems that generates meaningful notifications upon discovery of smarthome appliances. In Q3, I led a team of 3 engineers on -REDACTED-. I also became an Amazon Security Certifier, performed well in Amazon Machine Learning University. In Q4, I continued to work on -REDACTED- and also mentored an intern, who eventually ended up joining Amazon!
• In total, I've administered more than 50 interviews, contributed to Amazon-wide internal projects while maintaining high standards of design, security and implementation practices.

Jul 2019 – Oct 2020 · 1 yr 3 mos · Seattle, Washington

Feb 2016 – Jan 2017 · 11 mos · Greater New York City Area

• At Genius, I worked as a full stack engineer on Ruby on Rails/Angular JS stack. I was involved in making appropriate model changes, API changes, view changes and ensuring authorization controls. My main contribution was writing an internal framework built on top of Angular and service controllers that made building views with pagination highly efficient to develop. I was also involved in debugging hard-to-spot issues that involve synchronization of data.

May 2015 – Dec 2015 · 7 mos · Greater Los Angeles Area

• At Activision, I worked on building a generic recommendation engine using Spark (Scala). I achieved this after reading research papers, and worked on model changes to add regularization parameters to collaborative filtering techniques. I also forked a version of MLLib and made appropriate changes to build the recommendation engine.

Jul 2012 – Jul 2014 · 2 yrs

• I joined InMobi as an SDET; and shifted to development within a short span of time because of my consistently good performance.
• During my first year InMobi, I worked on the core Android SDK and also code owner of the JS/WinJS SDK offerings. These pieces of code run on approximately 800 million devices as seen on July 2014.
• After a year at InMobi working on the SDK, I decided to explore server side engineering. I got a chance to work with the geographical services team. I helped develop strategies and algorithms which resulted in throttled performance upto 8x-10x. This involved heavy pre-processing to eliminate realtime processing. This also involved using efficient data structures, using better indexing strategies and also by having data-sharding.

Jul 2012 – Nov 2012 · 4 mos

• As an SDET, my main contribution to InMobi was building non-trivial automation framework for black-box testing of android code. I also developed the Eclipse Plugin for InMobi and was also awarded for the same for it's ease of use.

Feb 2012 – May 2012 · 3 mos · Bangalore

• I developed framework for testing the InMobi SDK for Windows Phone.