Team · Lead, Accuracy & Macros Benchmarks (CCS-ACC, CCS-MAC)
Marcus Chen
AI Evaluation Lead
“Eval harnesses are how you tell a good AI feature from a great demo. We test the same prompts a real user would type, with the same edge cases.”
Who is Marcus Chen?
Marcus owns two of the six scoring dimensions: accuracy (25%) and macros (15%). Before joining The Calorie Counter Standard in 2023 he spent three years building eval harnesses at a Toronto applied-AI startup, where he became fluent in the gap between a model that performs on a benchmark and one that performs on the messy real-world prompts a user actually types. He designed the 50-meal calibrated reference set used by CCS-ACC and the macro-extraction test suite used by CCS-MAC, and he runs the weekly regression sweep against the nine apps. Marcus is also the person who decides whether an LLM update to one of the photo-logging apps justifies an out-of-cycle re-benchmark. He volunteers as a mentor for a Vancouver coding bootcamp for newcomers.
What does Chen specialise in?
- Eval harness design
- Prompt-robustness testing
- Voice and chat-logging accuracy
- AI coaching benchmarks
Where did Chen study?
- MSc, Machine Learning — University of British Columbia (2018)
- BSc, Computer Science — University of Waterloo (2016)
What did Chen do before The Calorie Counter Standard?
- Senior ML Engineer, evaluations group at a Toronto-based applied-AI startup (2020–2023)
- Software Engineer, recommendation systems, e-commerce platform (2018–2020)
What has Chen published?
- Chen M., Pearce J. "Reproducible accuracy benchmarks for consumer health AI." Internal preprint, 2024.
Which CCS calorie counter protocols does Chen own?
Marcus owns the following CCS scoring protocols. See the methodology page for what each measures.
CCS-ACCCCS-MAC
Reviews by Marcus
- Welling Review — 90.7/100
- MacroFactor Review — 74.8/100
How do you contact Chen?
For data requests on the accuracy or macro benchmarks, email research@calorie-counters.com.