The people behind Writence
A small team building for a large community of non-native English writers.
We are a remote-first group of engineers, designers, and researchers. Most of us write English as a second language, which directly shapes how we build the product.
THE TEAM
Who is building this
Small enough that everyone talks to writers; deliberate enough that everything ships with the mission in mind.
Alex
Founder & Engineering
Founded Writence after research into AI-detection bias against non-native English writing. Writes English as a second language.
Maya
Product Design
Shapes how the tools feel — calm, legible, and out of the writer’s way. Believes good design is mostly restraint.
Lena
Research
Tracks the literature on ESL writing and detection bias, and ensures product claims stay aligned with what the evidence supports.
Daniel
Engineering
Builds the shared engine behind Diglot and Copyeditor — the translation, drafting, and authorship layers.
Sofia
ML / NLP
Works on the language and model routing that make second-language writing feel native without flattening the writer’s voice.
Tomas
Support & Operations
Works most closely with the people who use the products, and turns writer feedback into the next fix.
HOW WE WORK
What it is like to build here
Remote-first
Distributed by default, async by habit. We hire for craft and judgment, not a time zone or a postcode.
Bilingual by experience
Most of us write English as a second language. We build the workflow we wished existed when we were the user.
Evidence over opinion
When we disagree, we defer to the research. The mission is grounded in published studies, and so are our product decisions.
We are hiring
Always glad to meet strong candidates
We do not open roles often, but we are always open to engineers, designers, and researchers who care about getting this right. If that describes you, the careers page is the place to start.
Building for
non-native English writers.
Review the open roles and our working process, or explore the products first.
Evaluating Perplexity Bias in Academic Detection
Quantitative assessments demonstrate that standardized models consistently mistake concise, rule-based second-language phrasing for synthetic generation.
The protocol generates a cryptographic signature confirming the record’s integrity, without exposing unpublished findings.