The most valuable resource in a university department isn't grant money or lab space. It's a professor's Sunday afternoon. Max Bohun and Aman Garg, both Cornell undergraduates, spent two years watching their instructors lose those afternoons to a mountain of handwritten problem sets and essays. Their startup, GradeWiz, is a simple proposition: let a computer do the first pass [Cornell Chronicle, February 2025].
It’s an AI grading assistant that uses computer vision to parse scribbled calculus proofs, physics diagrams, and biology sketches, alongside more standard PDFs and essays. The founders claim it cuts grading time by 60-80% and returns detailed feedback to students by the next day [GradeWiz site]. The wedge is straightforward. No district-wide procurement, no seven-figure sales cycle. Just a professor, drowning in papers, willing to try something that promises a weekend back.
The pilot wedge
GradeWiz’s early traction is a classic campus story. The founders, participants in Cornell’s eLab accelerator in 2024, started by solving a problem they saw every day. They piloted the tool with educators at their own university and a handful of others, including Pennsylvania State University and California Polytechnic State University [Cornell Chronicle, February 2025]. The reported metrics from these pilots form the core of their pitch: over 30,000 submissions graded, with 97% accuracy and less than 3% requiring manual intervention [Cornell Daily Sun, September 2025]. For a pre-seed company, that’s a compelling amount of real-world validation. It’s also the kind of data that got them into Y Combinator’s Winter 2025 batch and a $500,000 pre-seed round [Crunchbase, March 2025].
The unit economics of a saved hour
The business model, for now, appears to be a straightforward SaaS play aimed at individual educators or departments. The real unit economics here aren’t in the monthly subscription fee, which remains undisclosed. They’re in the hours saved. GradeWiz isn’t just selling software; it’s selling capacity back into a system running at a deficit.
The scaling cliff
Of course, moving from a dozen friendly university pilots to a scalable, defensible business is a different kind of problem set. The risks are well-defined.
- The founder gap. Bohun and Garg are sharp undergraduates with deep context on the user problem, but they have no public record of scaling a B2B SaaS company.
- The accuracy asymptote. A 97% accuracy rate in controlled pilots is impressive. Maintaining that rate across hundreds of institutions is the hard part.
- The competitor vacuum. The verified facts show no named direct competitors, which is either a massive greenfield opportunity or a sign that the market is historically difficult to crack.
The company’s immediate goal is to convert its pilot footprint into paying contracts and expand beyond the initial eight colleges [Y Combinator on X, 2025]. The next twelve months will test whether a product born in a Cornell dorm can build the sales and support infrastructure required to become a real company.