SharpScholar

Higher education has a feedback problem. Professors assign work, students submit it, and then weeks pass. By the time students get papers back, they've moved on mentally. The feedback, even when it's good, arrives too late to change behavior.

SharpScholar was a writing assistance platform. Students would write inside our tool, and as they wrote, they'd get real-time feedback—suggestions, questions, flags for common issues. The shift was from summative feedback (here's what you did wrong) to formative feedback (here's how to improve right now).

My co-founder and I were living the problem ourselves. We watched classmates struggle with writing—not because they couldn't write, but because they'd never learned how to write academically.

We interviewed professors and heard that they spent enormous time on feedback that students barely read. Meanwhile, students wanted help earlier in the process, not after they'd submitted. The spark was realizing that software could provide the early feedback that professors couldn't scale.

The core was a web-based text editor with a feedback engine. As you typed, the system analyzed your text for:

• Structural issues: Missing thesis statement, weak topic sentences, no transitions
• Argumentative logic: Claims without evidence, evidence without analysis
• Academic conventions: Citation format, formal tone, hedging language
• Common errors: Dangling modifiers, passive voice overuse, wordiness

Under the hood, we had a mix of rule-based systems and early NLP models. This was 2014-2015, pre-transformer era. We used dependency parsing, part-of-speech tagging, and hand-crafted heuristics.

The engineering challenge was making it real-time. Feedback had to appear as you wrote without disrupting the flow. We spent significant time on latency optimization.

We wanted credibility with academics, and in academia, that means research. We partnered with professors to run controlled studies—some students used SharpScholar, some didn't, we compared outcomes.

We published two studies showing improved writing quality for SharpScholar users. The effect sizes weren't massive, but they were statistically significant. That research became a key sales tool and fed back into product development.

Universities are notoriously slow to adopt new technology. Budget cycles are annual. Decision-making involves committees. Professors are autonomous and skeptical.

Our approach was bottom-up. Instead of selling to the institution, we sold to individual professors who could adopt tools for their courses. We'd find professors frustrated with the feedback problem, offer free pilots, and use their success to expand.

The RFP wins came later. Once we had 5-10 professors at a university using SharpScholar, administration noticed. Then we could talk about site licenses.

Credibility and flexibility. U of T had specific requirements around data privacy (student data had to stay in Canada), LMS integration, and accessibility compliance.

We built Canadian data residency specifically for the RFP. We did the accessibility audit ourselves. When they had concerns, we responded the same day. Larger competitors treated U of T as one of many deals; we treated it as the deal.

The metrics evolved as we learned:

Activation: What percentage of students who created an account actually used the tool for an assignment?

Usage depth: How many feedback suggestions did students see and act on? Surface engagement wasn't valuable; deep engagement was.

Retention: Were professors bringing SharpScholar back semester after semester?

Academic outcomes: Did SharpScholar users get better grades? Hardest to measure but most important for our thesis.

We built dashboards that professors could see for their classes. Usage data became a teaching tool—professors could see where students struggled.

Weekly sprints with demos every Friday. The demo discipline forced us to ship something visible every week. Monthly reviews on customer feedback and metrics. Quarterly strategy check-ins.

We were three co-founders wearing every hat. I leaned toward product and business development; co-founders leaned toward engineering. The lines were blurry. What I'd tell first-time founders now: specialize earlier than you think.

We'd hit a ceiling. EdTech is a brutal market. Sales cycles are long, budgets are small, and growth was slower than we needed to attract real funding. I wasn't sure I wanted to spend the next five years in that fight.

Shopify offered scale, mentorship, and the chance to learn from people who'd built massive products. The APM program was designed for people like me—startup experience but wanting to level up PM craft.

Was there ego involved in going from "founder" to "APM"? Sure. But title meant less than learning. And Shopify in 2016 was an obviously special company.

Customer obsession. When you're a startup, you can't afford to build the wrong thing. That intensity—constant customer conversations, data analysis, willingness to pivot—carried into everything I've done since.

Resourcefulness. At a startup, you solve problems with whatever's available. That mentality made me effective at Shopify, especially early when I'd do unglamorous work.

And the operating rhythms—weekly sprints, dashboards, structured reviews—I brought all of that. The discipline of running a tight operation translated.