Christopher Dewan

Software Engineer, Technical Leader, and Team Builder

Christopher Dewan

http://m3rlin.net
Chris.Dewan@m3rlin.net
Chiyoda-ku, Tokyo Japan

I have strengths in engineering, people and product management, and conflict resolution. I have a track record of working across teams, domains and oceans, with a strong focus on engineering/research collaboration. I'm looking for great coworkers, supportive management, and big technical challenges.


This resume is open source code.
Check it out on GitHub!

Experience

Meta Japan

Software Engineer, Instagram Relevance

Tokyo, Japan
Led successful major initiatives around reliability and user control in Instagram Recommendations in a cross-Pacific remote role.
  • Developed a new novel reliability metric for Instagram recommendation serving that was so simple and successful that it has become one of the primary reliability metrics across all of Instagram.
  • Led the engineering effort to build Instagram Recommendations Reset which was a very challenging technical project delivered on schedule and well under initial estimates of team size. I built the team, designed the architecture, set the schedule, drove the execution, and built several of the most difficult components. All of my teammates on this project were in the United States.
July 2022 — Present

Meta Platforms (formerly Facebook)

Software Engineer, Meta AI

Menlo Park, CA, USA
Research Engineer, FAIR
Led the engineering-side efforts in a large language model research team in Meta AI's FAIR organization.
  • Restructured how we set goals and execute on them, getting the whole team working on common objectives and collaborating to achieve them.
  • Negotiated and executed a large refactor of the Metaseq codebase, allowing it to be open sourced as part of the Open Pretrained Transformers (OPT) model and paper (https://arxiv.org/abs/2205.01068) release without disrupting internal research.
  • Helped organize the transition of the OPT effort into following LLM efforts, which were quite successful.
February 2022 — June 2022
Technical Lead for MultiRay, TextRay, and PostRay, Facebook AI
As overall technical leader for content understanding. I was the responsible authority for the strategy, technical architecture, and execution of our team spanning multiple key computer vision and natural language processing (NLP) problems across all of Facebook (Meta). This project, called MultiRay, changed how NLP and multimodal understanding was done cross the company.
  • Wrote the business case for MultiRay, (https://ai.meta.com/blog/multiray-large-scale-AI-models/) a centralized state of the art (SOTA) text understanding service at Facebook. Built the engineering/research partnerships, designed the system architecture, and implemented several of the most complex components. This was the flagship project of our team and the premier text understanding system at Facebook. We delivered it into production despite a company wide capacity crunch and delivered significant business impact working as #oneteam across organizational boundaries.
  • Partnered across organizations to build the case for a centralized SOTA multi-modal content understanding system. Led the combined research and engineering work that brought it to production as #oneteam.
  • Drove the technical and business strategy for a new team that centralized content understanding AI services used to understand both language and images across Facebook and Instagram.
  • Drove workload definitions and initial business case for large scale research training clusters at Facebook
  • When COVID-19 Caused a simultaneous jump in internet usage and global supply chain challenges I drove Facebook AI's efficiency efforts far beyond the target baseline and we delivered big wins.
  • Mentor to 14 simultaneous mentees, including IC's and managers across multiple teams. I helped them achieve career goals including promotions, project successes, and team changes.
February 2020 — February 2022
Tech Lead Manager for Facebook AI NLP Infrastructure
Combined role of both people management and technical leadership of the NLP Infrastructure team.
  • Expanded the team's scope beyond inference, building a collaboration with with the FB Assistant Language team to build the open source PyText training platform to bring PyTorch based NLP models to all of Facebook.
  • Led the push to bring new hardware capabilities into FB in a very short timeframe. This unblocked the research that resulted in the RoBERTa and XLM-R models.
  • Hired and grew my team from 4 to 8, and supported the growth of individuals on the team.
  • Drove a broad effort across Facebook AI to improve engineering standards.
  • Built the business case for a crash effort to bring GPUs to production for inference for the first time at Facebook and provided management cover for this controversial project whose success changed FB AI Hardware Strategy.
  • Recognized that by doing two jobs I was a bottleneck for the team growth, as I couldn't support enough people to match our needs. I chose to split my role into two and reverted back to Technical Lead.
July 2018 — February 2020
Facebook AI NLP Infrastructure Technical Lead
As engineering lead for NLP Infrastructure, getting research into production was my job.
  • Built a generic auto-scaling microservice system for serving many ML models efficiently. This was extended beyond NLP and took over most non-ads ML inference workloads at Facebook.
  • Greatly expanded the scale and efficiency of Translation inference at Facebook.
  • Supported my manager and helped hire and grow new talent within the team.
September 2017 — June 2018
Translation Infrastructure Engineer
As the first infra engineer on Facebook Translation, I made our system more scalable and reliable
  • Drove the inference infrastructure migration from statistical to neural machine translation at FB.
  • Rearchitected translation inference to significantly increase reliability.
  • Built the hot-fix system used to respond to bad translations at Facebook for Neural Machine Translation
November 2015 — September 2017

Raytheon

Senior Multi-Disciplined Engineer

El Segundo, CA, USA
Adviser to Software Center Management
Led effort to improve hiring and retention for software engineers. I received an award for this work.
  • Created a tailored new interview loop focused on relevant skills with expanded use of technical questions.
  • Organized an effort to ensure new hires would have real work, improving morale and retention.
January 2015 — September 2015
VIIRS Ground Support Equipment Responsible Engineering Authority
I was the engineering owner of the ground station and spacecraft simulators used to test VIIRS (Visible Infrared Imaging Radiometer Suite), a calibrated spaceborne weather sensor. My upgrades reduced errors during the test program by 80%. I received 3 personal and 7 team Mission Success Awards for this work.
  • Made the case for, designed, and executed a rewrite of the simulator software. (~40,000 lines of C++)
  • Designed and built an adapter board so inexpensive commercial hardware could be used in test.
  • Redesigned the simulator for SpaceWire connectivity to support VIIRS upgrades.
  • Built and maintained a data aggregation system which cut analysis time from hours to minutes.
  • Provided 24/7 on call support for months of continuous testing.
August 2012 — September 2015
Raytheon Capital Test Station: Electrical and Software Architect
Software/hardware design of a test station offering new capabilities to customer programs and internal R&D. I received a Raytheon Innovation Award for the novice configuration database I designed.
  • Gathered electrical and software requirements and designed an overall architecture, selected the hardware, and did detailed hardware design, both electrical and mechanical
  • Designed and built the software stack, including a web app and API for configuration data and telemetry, along with a client library integrated into LabVIEW, our test software platform. (~10,000 lines of C# and HTML)
April 2011 — August 2012
Test Systems Engineer
Designed Test apparatus and components, supported test programs and R&D efforts.
June 2010 — April 2011

NTT Communication Science Laboratories

Summer Intern

Atsugi-shi, Kanagawa, Japan
Worked in the vision group of the NTT Behavioral Research Lab in the NTT Atsugi R&D Center. Studied how reflections affect human visual inference of materials and illuminants.
  • Experimented with machine vision techniques for manipulating specular and matter reflection in an image.
  • Designed and built an apparatus to separate matte and specular reflection in a scene using polarization.
June 2009 — August 2009

Caltech Physics Department

Laboratory Teaching Assistant

Pasadena, CA, USA
October 2008 — December 2009

Languages

  • English — Native speaker

  • Japanese — Professional Working Level. Passed JLPT N1

Education

Caltech — BS, Physics

October 2006 — June 2010

Skills

Programming

  • Major Work C++
  • Major Work Python
  • Major Work PyTorch
  • Minor Work C#
  • Minor Work Javascript
  • Minor Work HTML
  • Minor Work CSS

Engineering

  • Major Work Distributed Systems
  • Major Work Unit Testing
  • Major Work Continuous Integration
  • Major Work Analytics
  • Major Work AI Hardware Strategy

Leadership

  • Major Work People Management
  • Major Work Mentorship
  • Major Work Cross Functional Alignment
  • Major Work Performance Review

Interests

Photography

  • Landscape
  • Product
  • Motorsports

Reading

  • Current Events
  • History
  • Economics
  • Manga
  • Biography

Technology

  • Computer Hardware
  • Flight Simulators
  • Virtual Reality