Machine learning researcher at Google BrainCurrently focused on research for parameter-efficient methods of control for large pretrained language models and explicit methods to induce generalization to new tasks.

AI Resident at Google Brain working on Natural Language Processing and Understanding.

Specializing in using Deep Learning for NLP. Researching new training methods and model architectures, building client models for production, and creating infrastructure that enables distributed, cloud-native, training and horizontally scalable model serving. I maintain Mead-Baseline—our open-source deep-learning toolkit. It is the go to path for all deep learning work (research and production) within the company. I have created performant implementations of various complex neural network architectures including a Conditional Random Field and Beam Search.Designed the label space and annotation guidelines for the Natural Language Understanding module of a customer self-service dialogue system in the technical support domain. My design focused on general intents, complex entities, and relations to cover diverse and complex conversational domain.Created our cloud-native, deep-learning model training platform. Out platform is a DAG engine powered by kubernetes that can execute any pipeline of tasks that can be run inside of docker containers. Our platform automatically parallelizes non dependent nodes in the DAG (allowing for HPO) and can distribute a single step over multiple GPUs. The platform enables building mutli-step pipelines that can take a model from raw training data to a model that is ready for production.Created and Maintain our deep-learning model serving infrastructure. Backed by TensorFlow Serving, we enable rich NLU via cascading calls to deep learning models. The server itself and TensorFlow Serving are deployed via Kubernetes and are powering several production dialogue systems.Created a wide range of Machine Learning models powering client-facing production models. The architectures and tasks range from ConvNets for classification (used in intent detection), bLSTM-CRF taggers for general purpose tagging (NER, POS, Chunking) and client-specific slot filling, to Transformer-based seq2seq models used to automatically suggest agent responses.

Used machine learning, natural language processing, computer vision, and cutting edge deep neural networks to power improve heuristic based existing features and create new features. Designed and Implemented our model serving infrastructure that processed 200 emails a day. Provided technical leadership to the data science team.Trove would detect and surface sentences in emails that warranted a response to help track emails that were important to reply too. This was originally done via a tangled mess of regular expressions. Using unsupervised methods and iterative refinement I bootstrapped a dataset of these sentences. Using a ConvNet I was able to create a system better than the regex. The ConvNet was also able to account for semantic meaning and we refined the product so that it highlights requests that can be handled via email not just any ask. The results of this model became a staple feature in the annotated email social graph that trove created.Used a neural ranking model to preform coreference resolution by linking anaphors to their antecedents. By replacing pronominal mentions in the question snippets extracted from emails we were able to provide context to the users.Used lexical features from the email as well as connectivity information in the email social graph to identify bot accounts. We used a broad definition of bots to include things like newsletters and automated email from things like github. We used these classification labels to help filter search results.

Specialized in Voice Recognition and User Information Systems.Prototyped and Vetted Dragon Drive Link, an on-board real-time driver information and entertainment application similar to Android Auto.Research, designed, and implemented a system to optimize Voice Recognition used in Production Mazda cars. It dynamically decides if it should display a menu of disambiguation options based on the users past interactions with the system. We patented with system (US 9984688B2)