Jan 2025 – Present · 1 yr 4 mos

• Operator Collective is a venture fund and dream team community of operator LPs. OpCo brings together 200+ of tech’s most exceptional executives from diverse backgrounds to invest in and supercharge the next generation of enterprise tech.

Nov 2021 – Present · 4 yrs 6 mos

Jun 2021 – Jun 2025 · 4 yrs

• I lead Discovery & Promotion – Machine Learning, Data Science & Engineering at Netflix, where we push the boundaries of Generative AI, LLMs, VLMs, and multimodal learning (text, image, audio, and video) to enhance content promotion, discovery and fan engagement.
• Our teams build cutting-edge ML-powered creative tools—empowering artists and creators to generate compelling artwork, synopses, and videos. We leverage contextual bandits and multimodal AI to personalize and optimize Netflix’s recommendations, marketing campaigns, and fan experiences on Tudum.
• Beyond innovation in personalization and promotion, my teams spearhead R&D initiatives shaping the future of content production—from multimodal content understanding of movies, TV, and games to large-scale, AI-powered operations. We’re also making bets on Live Content & GenAI to redefine how stories come to life.
• At the heart of it, we build ML & AI products that fuel fandom, enhance creativity, and deliver joy to millions.

Jun 2016 – Jun 2021 · 5 yrs · San Francisco Bay Area

• I led the organization that revolutionized iOS; making it intelligent, timely, relevant and focussed.
• Our proactive suggestions accelerate your workflow by predicting apps you may want to launch next, actions within apps you may want to perform and people you want to interact with based on the context and adapt them to iOS UIs including lock screen, Spotlight, the newly designed smart homescreen and the App library.
• We employ a host of large scale and on-device personalization algorithms to surprise and delight with personalized contextual suggestions and deploy them to 1.65+ billion devices!

Jan 2014 – Jan 2016 · 2 yrs · Mountain View

• I led Next gen ML efforts in Intuit’s Core Technology Organization (CTO). My focus was on:
• Image Information Retrieval: Automatically extract form information from images using computer vision, machine learning, image processing and OCR techniques. This powers for example, our feature where snapping a photo of your W2 extracts all the tax information.
• Vertical Search: Optimization of ranking and relevance metrics by using careful feature engineering, incorporating user feedback, machine learned ranking / learning-to-rank techniques. We supported the faceted enterprise search application in LiveCommunity (In-house Q&A community forum for everything tax) and helped people get answers to complex tax questions.

Jan 2014 – Jan 2014 · 0 mo · San Francisco Bay Area

• Research, modeling and prototyping solutions to discover users contributing to spam, unsolicited behavior and content. Built classifiers to identify bot-like behavior, suspicious registrations and unwanted guests to improve user experience on the social media website.

Jan 2012 – Jan 2014 · 2 yrs · Greater Madison Area

• The R&D team I was part of at Networked Insights is responsible for delivering powerful algorithms for large scale data mining on unstructured social media data for visualizing audience behavior using semantic word embeddings, automatically infer demographic attributes using NLP and Machine Learning, categorizing unstructured data into marketing taxonomies and the scientific aspects of various products to help marketers with audience insights. I led and worked end-to-end on several products including:
• Geolocation inference: Automatically infer home city from short social media posts using deep learned language representation and Machine learning for hierarchical classification.
• Predicting box-office performance: Predict box office performance of movies from social media data across multiple audiences using time series analysis and dynamic Bayesian networks.
• Labeling social media posts: Developed a suite of products for labeling social media posts. This involved exploiting high information density features like CRF based Entity extraction, phrase extraction using information theoretic criteria and topic models for discovering themes. In addition, distributed word representations were used for capturing semantics.

Apr 2010 – May 2011 · 1 yr 1 mo

• While my expertise is in machine learning, I have experience developing database-driven web applications as a full-stack developer in Space Science and Engineering Center and Department of Agricultural and Applied Economics. Technologies used include Ruby on Rails, python, MySQL and CodeIgniter (PHP MVC Framework).

Jan 2010 – Jan 2012 · 2 yrs

• Worked on Statistical Machine translation, Bayesian models and approximate inference using MCMC.
• Translation of poetic texts involves serious alignment problems due to missing passages, agglutinative language and large differences in syntax, which calls for joint word and sentence alignments for learning translations. I detected gaps using a structured perceptron and developed a probabilistic extension of dynamic time warping algorithm for word and sentence alignments.
• Adaptive assessment system for second language acquisition: Developed an adaptive test based learning system to help Korean students learn English. The assessment system scored the semantic and syntactic correctness of the learner and adaptively adjusted the difficulty of succeeding problems using Q-learning.