- Joined
- Aug 2024
- How they joined
- Directly
- Before Cognition
- Scale AIStaff Software Engineer
- Status
- Current
- Joined Cognition (maker of Devin and Windsurf) as a Software Engineer in August 2024, after seven years at Scale AI; the move came about a month after leaving Scale in July 2024.Aug 2024
- At Scale AI (June 2017 - July 2024) rose to Staff Software Engineer and was 'Infrastructure lead and generalist' and technical lead of the Platform team, overseeing Scale's core shared infrastructure, libraries and services as the team grew from 3 to 12 engineers.Jun 2017
- Member of the 2013 US Physics Team at the International Physics Olympiad in Copenhagen: won gold, placed 8th overall, and earned the highest experimental score of the competition. Teammates on the five-person traveling team were Kevin Zhou (gold, 5th), Jeffrey Yan (gold, 24th), Jeffrey Cai (silver) and Samuel Zbarsky (silver), coached by Paul Stanley.Jul 2013
- Posted on LinkedIn on Scale's Series C (Aug 2019), word for word: "It's been a great journey with Scale AI for the past two years, from a small 10-person company to a unicorn!" and "We are building infrastructure to accelerate the development of AI applications. We've only just begun by solving the data labeling problem." He joined when Scale had about 10 people.Aug 5, 2019
- At Scale, architected the control plane and management procedures of the Scale Generative AI Platform, described as Scale's first product suite deployed into commercial customer environments; this infra-for-customer-environments background lines up with enterprise deployment work at Cognition.
- Commented on an Aug 2025 LinkedIn post claiming Scale AI's original team got 'zero upside' after Meta's deal, as 'Software Engineer at Cognition': "It is false to say the original team received zero upside at scale." (comment text as indexed by search), a public defense of his former employer one year into Cognition.Aug 2025
- Named to Exceptional Capital's Exceptional 100 (2025) list under Cognition: 'Calvin is a software engineer at Cognition.'2025
- Attended the 2013 US Physics Team training camp (University of Maryland) with camp-mates including Zachary Bogorad, Kevin Zhu, Grace Lin, Kevin Fei, Matthew Huang, Eric Chen, Jonathan Tidor, Edward Park, Owen Gray, Vickie Ye, Rahul Sridhar, Mike Winer, Xianglong Ni and Lawrence Sun; coaches Lucy Chen, Marianna Mao, David Fallest and Andrew Lin.2013
- Public headline says "Software Engineer at Cognition".
- Drove service-oriented development at Scale by creating service and library templates used by all engineering teams, with 30 services deployed per his profile.
- Undergraduate researcher on MIT's dReal SMT solver project, developing branching strategies; later first author (C. Huang) on 'Evaluating Branching Heuristics in Interval Constraint Propagation for Satisfiability' (2019) with Damien Zufferey.2019
- Undergraduate researcher in MIT CSAIL's Spoken Language Systems group (Jan-May 2015) on the Nutrition System, improving semantic labelling of foods with word-vector features; collaborators named in CSAIL's report include Mandy Korpusik, Rachael Naphtal, Patricia Saylor and James Glass.2015
Infrastructure lead and generalist. Technical lead of the Platform team. Oversaw the development and maintenance of Scale’s core shared infrastructure, libraries, and services. Saw the team grow from 3 to 12 engineers. • Collaborated with other engineering teams to help them make scalable architectural decisions, and to gatherfeedback to shape the roadmap of Platform team’s supporting services. Architected the control plane and management procedures of the Scale Generative AI Platform, Scale’s first product suite deployed into commercial customer environments. Drove the adoption of service-oriented development by creating service and library templates, drafting and sharing best practices with the engineering organization. Currently used by all engineering teams at Scale, with 30 services deployed and growing.
Conducted an investigation on techniques to improve search infrastructure reliability, and redesigned the search indexing pipeline to be simpler, more reliable, and easier to use. Created tools to help automatically transition the codebase to Python 3.
Designed, developed, and deployed new search suggestions system for Google Fiber TV. Reduced number of keystrokes required for top suggestions by 45%.
Helped develop Nutrition System, which displays and tracks nutrition information from a spoken meal description. Improved semantic labelling of foods and food properties using properties of word vectors as features.
Developed and prototyped automated testing solutions for various components of the MM3 energy monitor.
Created utilities to generate and analyze solar simulations; analyzed simulation output and empirical data to improve simulation methodologies—submitted findings to Intel Science Talent Search
Researched and created heuristics to improve the accuracy of Semantic Role Labeling and studied the effects of using structured learning on the accuracy of SRL—received special award from the Association of Computing Machinery at the Synopsys Science Fair
