
Personalization @ OpenAI, previously at Microsoft AI, Inflection AI, and Meta
AI is analyzing your overall score…
Identifying your key strengths…
Evaluating your skill match against the job requirements…
Assessing your cultural and operational fit
Member of Technical Staff at OpenAI, working on applied research for personalization. I previously worked on post-training for in-house foundation models at Microsoft AI, and before that I was at Inflection AI, where I was a key contributor to building Pi from the ground up. 7+ years of Conversational AI experience, including at Microsoft Semantic Machines, Meta Conversational AI (on Portal and RayBan glasses tech), and Siri.
Stanford University
Master’s Degree, Computer Science
January 1, 2015 – January 1, 2017
Columbia University
Bachelor's Degree, Computer Science, Linguistics
January 1, 2011 – January 1, 2014
OpenAI
Member of Technical Staff
August 1, 2025 – Present
Microsoft
Member of Technical Staff
March 1, 2024 – August 1, 2025
Inflection AI
Member of Technical Staff
September 1, 2022 – March 1, 2024
Microsoft
Principal Researcher
September 1, 2019 – September 1, 2022
Machine Learning Engineer
August 1, 2017 – September 1, 2019
Menlo Park, California
Software Engineering Intern
June 1, 2016 – September 1, 2016
San Francisco Bay Area
Apple
Machine Learning and Natural Language Engineer, Siri
February 1, 2015 – July 1, 2015
Cupertino, CA
Apple
Siri Intern
June 1, 2013 – August 1, 2013
Cupertino, CA
Columbia University - Computer Science Department
Head Teaching Assistant
September 1, 2012 – December 1, 2014
Greater New York City Area
ClassWish.org
Intern
February 1, 2012 – October 1, 2012
Greater New York City Area
Hacking Violent Extremism Online
September 1, 2016 – Present
Collaborative project with State Department to apply innovation principles and technology to deeply understand the Countering Violent Extremism space and design solutions.
Multi-task Multi-modal Deep Neural Networks for Predicting Transcription Factor Binding Sites
September 1, 2016 – Present
As part of the DREAM challenge (https://www.synapse.org/#!Synapse:syn6131484) experiment with various multi-task multi-modal neural network architectures to accurately predict genomic binding sites for various transcription factors (TFs) in different cell types. The basic problem is to make these predictions for each TF using genomic sequence data (which stays constant across cell types) and chromatin accessibility information (which differs from one cell type to another). Using a multitask architecture allows us to train the network for multiple TFs at once, allowing the network to better learn the features that predict binding across TFs.
Recurrent Neural Networks for Understanding Music
April 1, 2016 – June 1, 2016
Implemented recurrent and convolutional neural network-based architectures to determine whether two songs are similar or not based on their lyrics or audio spectrograms.
Convolutional Neural Networks for Outcome Classification in Sports Videos
January 1, 2016 – March 1, 2016
Implemented and explored the performance of various convolutional neural network (CNN) based architectures on the task of play-by-play outcome classification for sports videos (specifically for cricket).
Deep Learning for Dialogue Agents
September 1, 2015 – Present
- Created a statistical natural language understanding system using RNNs that understands and responds to commands to arrange blocks in a virtual block world. - Currently working on modeling agents that can participate in collaborative goal-oriented dialogue-based tasks.
Memorability
March 1, 2014 – May 1, 2014
Identifying whether a given utterance is memorable or not has tremendous utility in advertising and education, among other domains. We describe an SVM classifier that uses linguistic cues to predict whether an utterance will be memorable or not. The feature set is based on existing psycholinguistic research, and we describe ways to computationally extract those features.The idea behind this work is to come up with features which are capable of identifying memorable texts across multiple domains.
WordsEye Linguistics Tool (WELT)
September 1, 2013 – December 1, 2014
Advisors: Dr. Julia Hirschberg, Bob Coyne, Morgan Ulinski WELT is a tool for fieldwork in linguistics. We are creating an interface to map from the syntactic representation of a sentence in the target language to a structure that can be used by WordsEye to generate a scene in that language. I also assist in the elicitation of the target language, Arrernte, by creating 3-D images. The information from these elicitation sessions is in turn used to build a grammar for the language that WordsEye can use.
Tweet-to-Scene Conversion
September 1, 2012 – Present
Advisors: Dr. Julia Hirschberg (Chair, Computer Science Department), Bob Coyne WordsEye is a system, developed at Columbia University's Speech Lab, that converts descriptive text into representative 3D images. Our goal is to develop a system that processes and understands information represented by tweets from Twitter, analyzing them for semantics. This includes gauging the emotional content of the tweets, handling Internet slang terms and popular idioms, and filtering for current trends in the content of tweets. The system uses pattern recognition and part-of-speech taggers to identify these elements in observed tweets in a data-driven fashion. The goal is to convert a tweet into a meaningful representation that WordsEye can directly interpret to create an appropriate scene.
Cultural Fit Analysis
The candidate has a strong background in research-oriented and product-focused roles within large tech companies, indicating a fit for fast-paced, innovative environments. Their diverse personal projects, ranging from linguistics tools to deep learning applications in various domains, suggest intellectual curiosity and a proactive approach to learning and problem-solving. The progression through roles at top-tier companies like OpenAI, Microsoft, and Facebook aligns with a high-performance culture. However, the target role is 'Backend Engineer', and while their ML/NLP background is strong, there's limited explicit mention of traditional backend engineering skills (e.g., distributed systems, API design, database management, specific programming languages beyond Java for teaching). This might indicate a potential gap in direct alignment with a pure backend engineering role, though their ML engineering roles likely involved significant backend components.
Soft Skills & Operational Fit
The candidate's experience as a Head Teaching Assistant at Columbia University suggests strong communication and mentorship skills. Their project descriptions indicate an ability to collaborate on complex problems (e.g., Hacking Violent Extremism Online, WordsEye Linguistics Tool). The roles at multiple prominent companies imply adaptability and a strong work ethic. However, without psychometric test results, a deeper assessment of stress handling, work attitude, and team collaboration is not possible.