Entrepreneurship & Innovation
The same methods that anchor my research — field experiments, A/B testing, and applied machine learning — also shape how I build. I'm interested in ventures where empirical rigor and real classroom or platform constraints meet.
Co-Founder | Ednius
June 2024 – Present
"A student, a professor, and a problem worth solving."
Students wait weeks for feedback that's often too generic to act on. Educators spend 10–22 hours a week grading per course, and by the time responses arrive, the learning moment is gone. Ednius is an AI-assisted grading platform for universities and certification bodies that closes that gap — generating personalized, rubric-aligned feedback on open-ended student work while keeping the educator firmly in the loop on every approval.
To date the platform has processed 60,000+ submissions with 97% accuracy and delivers around 70% time savings for instructors across institutions in North America and Asia. Four principles anchor the product: educators retain final decision-making authority, student privacy is protected, feedback is prioritized over grades, and evaluation standards stay consistent across assignments and graders.
The product is built the way I run a study — every feature ships behind an A/B test, and design decisions are driven by field-experiment evidence rather than intuition.
Learn more or get in touch: www.ednius.com
Call for Research Collaboration
Ednius sits at an unusually rich intersection — generative AI, higher education, assessment, and human-in-the-loop decision systems — and it generates research questions faster than a single lab can answer them. If you're a researcher, doctoral student, or thoughtful practitioner working on any of the directions below (or something adjacent), I'd love to hear from you. I'm open to co-authored studies, field experiments run on the platform, and student projects.
- Trust calibration in human–AI grading. When does an instructor over-trust an LLM's suggested score, when do they under-trust it, and what interface cues push that calibration toward the truth? A natural setting for online experiments with real graders.
- Feedback that actually changes behavior. Not all feedback is acted on. Which structural features of AI-generated feedback (specificity, framing, timing relative to the learning moment) measurably improve subsequent student performance? An ideal field-experiment question on a live platform.
- Fairness and bias auditing in LLM-assisted assessment. Do LLM scoring patterns differ across student writing styles, English fluency, or disciplinary conventions in ways that traditional rubric analysis would miss? A timely question with policy and accreditation implications.
If any of this resonates — or if you have a research idea I haven't listed and want a real-world platform and dataset to test it on — please reach out. I'd genuinely love to talk.