Mar 2022 – Apr 2025 · 3 yrs 1 mo

• Leading top-class ML infrastructure engineers solving ML feature, dataset and tooling infrastructure problems at Pinterest.
• Large scale ML feature store, training dataset management and user event sequence systems supporting O(10k) features, O(100) Pb data, O(100m) QPS, and transformer based recommenders to bring inspiration to over 500M pinners across the globe.
• Internal fullstack ML tools supporting hundreds of ML engineers
• In charge of managing the company-wide ML infrastructure spend, strategizing hardware choices, and driving savings initiatives.

Jun 2017 – Apr 2022 · 4 yrs 10 mos

• Tech-lead and founding member of the ML Platform team.
• Built and scaled end-to-end ML infrastructure for O(100) ML engineers and O(100M) pinners.
• Spearheaded transitioning ML infra to support deep neural network architectures from classical ML algorithms.
• Key contributions: centralized feature store, company-wide ML data format standard, training frameworks, model deployment systems.

Nov 2014 – Aug 2016 · 1 yr 9 mos · Republic of Korea

• As the squad leader of the EMM Company HQ, I was responsible for the health, welfare, morale, and training of the 12 soldiers in my squad. I received ROK Army certified leadership and counseling training. I received 4 medals for excellence in combat training and Squad management.
• At the same time, I also held the responsibilities as a Automated Systems Administrator at my base. My primary job was to manage and operate the automated IT systems in the ROK Army, and support my regiment in its operations.

Jun 2014 – Sep 2014 · 3 mos · Stanford

• As an intern at the Stanford OpenEdx team, I served as one of the core contributors to the OpenEdx opensource project, a massively open online course (MOOC) provider. I worked along with developers from MIT, Harvard, Stanford, and other universities to develop scalable features for users around the world.
• I focused on Edx's backend python server development, and successfully completed self-driven projects in performance improvements and feature implementations on Edx's Course Management System.

Oct 2013 – Sep 2014 · 11 mos · Stanford

• As a part-time prototype engineer at Tackl, I was responsible for developing its crowd-sourced problem solving web platform. Tackl was a small startup in process for its initial funding when I was first introduced to the CEO to build its prototype service. I mainly worked with Ruby on Rails for its backend, and ported the web application onto Amazon EC2 and RDB instances for beta testing.

Jun 2013 – Sep 2014 · 1 yr 3 mos · Stanford

• As the co-founder and CTO of Deimos Defense, I was in charge of developing the world's first smart self defense spray. Deimos is a digitalized pepper spray device that equips onto smartphones via 3.5mm audio jack. When it is operated, it shoots pepper spray and activates the smartphone via audio signals to take pictures of the assaulter, gathers real-time crime information and shares it with the user's contact list.
• Deimos received $110,000 venture grant from the South Korean government and TIDE institute.
• I designed and developed the architecture for Deimos' backend server, and its mobile applications (Android, iOS). I also designed the hardware architecture and funcationality, which processes and interprets the sound information sent from the smartphone to operate the device. With this method of controlling the smartphone device via sound channel, I was able to reduce the initial production cost of Deimos by 90% compared to our initial bluetooth prototype.

Jun 2012 – Aug 2012 · 2 mos · Bangalore, India

• As an intern at Inmobi I worked on several projects such as product / api localization and training the sales employees about its automated ad serving algorithm