Keith Stevens, Developer in Hakuba, Nagano, Japan
Keith is available for hire
Hire Keith

Keith Stevens

Verified Expert  in Engineering

Machine Learning Developer

Hakuba, Nagano, Japan

Toptal member since August 17, 2022

Expertise
Machine LearningNatural Language Processing (NLP)LinuxGoogle CloudPythonJavaScriptReactC++JupyterDockerServerlessJavaGoogle BigtableGit
Bio

Keith builds natural language processing (NLP) powered products. He has worked on almost every part of Google Translate's stack from training models, managing training data, deploying models, and integrating models into user experiences. Keith believes it is essential to leverage user feedback for high-quality experiences.

Portfolio

Design To Be
Python, Serverless, Node.js, React, Hugging Face...
Google
C++, Python, JavaScript, Generative Pre-trained Transformers (GPT)...
Google
C++, JavaScript, Java, Python, Flume, Generative Pre-trained Transformers (GPT)...

Experience

Generative Pre-trained Transformers (GPT) - 12 yearsMachine Learning - 12 yearsNatural Language Processing (NLP) - 12 yearsC++ - 12 yearsMachine Translation - 9 yearsPython - 8 yearsJavaScript - 6 yearsReact - 2 years

Availability

Part-time

Preferred Environment

Linux, Jupyter, React, Docker, Python, JavaScript

The most amazing...

...feature I've launched is beta machine translation models for low resource languages powered by user contributed data.

Work Experience

CTO

2022 - 2022
Design To Be
  • Developed and deployed a prototype slack bot that summarized conversations in Slack and Figma to produce weekly reports.
  • Tested capabilities of several large language models and trained prototype text classifier models.
  • Investigated the current market for significant language model-based applications and planned the unique features to develop.
Technologies: Python, Serverless, Node.js, React, Hugging Face, Natural Language Processing (NLP), Generative Pre-trained Transformers (GPT)

Staff Software Engineer

2020 - 2022
Google
  • Developed a crowdsourcing platform that supported over 200 languages and gathered millions of high-quality sentence pairs.
  • Created fully automated pipelines that retrained and deployed machine translation models of various sizes when new data was available.
  • Managed three direct reports and replaced the server and client-side implementation of user-facing products with zero downtime.
Technologies: C++, Python, JavaScript, Generative Pre-trained Transformers (GPT), Natural Language Processing (NLP), Machine Translation, Google Cloud

Senior Software Engineer

2015 - 2020
Google
  • Created parallel translation data miners for extracting and filtering trillions of translated sentences from the web.
  • Developed data repositories for storing and managing hundreds of datasets covering over 100 languages and containing trillions of translations.
  • Set up internal tools for interactively exploring and debugging sequence-to-sequence machine translation models before their launch.
  • Launched two iterations of Translate Contribute to collect hundreds of thousands of translated words and phrases from volunteers.
Technologies: C++, JavaScript, Java, Python, Flume, Generative Pre-trained Transformers (GPT), Natural Language Processing (NLP), Machine Learning

Software Engineer

2012 - 2015
Google
  • Consolidated multiple licensed word and phrase level translation datasets into a single data pipeline and back-end server.
  • Developed user feedback features on Google Translate.
  • Redesigned and deployed core serving infrastructure for Google Translate to support multiple new clients in different product areas.
Technologies: C++, Python, JavaScript, Java, Google MapReduce, Google Bigtable

Teaching Assistant

2008 - 2012
University of California, Los Angeles
  • Reviewed course lectures with students once a week for two hours and detailing homework expectations.
  • Graded weekly assignments and reviewed answers with students.
  • Prepared and graded exams with the professor and other teaching assistants.
Technologies: C++, Python, Operating Systems, Git

Contribution to Google Translate

Created a feature to Google Translate's web app that allows users to contribute data to translate. I started the project from its initial launch in 2014 until 2022, when I left. I developed user interfaces, server side APIs, and data management pipelines. In the last two years, I also managed three team members that took over those responsibilities while I focused on using the gathered data to develop advanced data cleaning methods and deploy beta translation models.
2008 - 2012

Master's Degree in Computer Science

University of California, Los Angeles - Los Angeles, CA

2005 - 2008

Bachelor's Degree in Computer Science

University of California, Los Angeles - Los Angeles, CA

Libraries/APIs

React, Node.js

Tools

Jupyter, Flume, Git

Languages

Python, JavaScript, Embedded C, C++, Java

Platforms

Linux, Docker

Storage

Google Cloud, Google Bigtable

Other

Machine Learning, Natural Language Processing (NLP), Generative Pre-trained Transformers (GPT), Machine Translation, Operating Systems, Computational Linguistics, Programming Languages, Google MapReduce, Serverless, Hugging Face

Collaboration That Works

How to Work with Toptal

Toptal matches you directly with global industry experts from our network in hours—not weeks or months.

1

Share your needs

Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2

Choose your talent

Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3

Start your risk-free talent trial

Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.

Top talent is in high demand.

Start hiring