I have spent over a decade turning science into learning, first as a classroom teacher and science program director, now as an academic specialist producing curriculum-aligned lessons at scale. These four pieces show the problem each one solved and how I built it.
A desktop app that generates high-school chemistry worksheets with answer keys. A teacher picks topics, sets the options and gets a print-ready sheet in two shuffled versions, with a key that shows the computation and the answer. It covers twelve topics across moles, stoichiometry, measurement and gases.
Chemistry practice runs on numbers, and numbers make answer keys fragile. A mass-to-mass problem carries molar masses, rounding and significant figures through to one answer, and a student who rounds a molar mass differently reaches a neighbouring value. Handmade sheets get reused and their answers circulate, while fresh sets mean redoing the arithmetic. The design question was how to generate unlimited problems whose keys a teacher can defend, including when two reasonable routes end at different last digits.
High-school chemistry teachers who prepare practice sheets, quizzes and review sets, and the students who work them. Each sheet follows the teacher’s own conventions: atomic-weight precision, rounding rule, Avogadro’s number, STP conditions, gas constant and Kelvin offset. It matches what the class was taught.
Product owner, assessment designer and chemistry lead. I scoped the product, specified every chemistry rule, designed the topics, item types and answer-key standard, ordered the build, reviewed the generated items, and directed the AI-assisted development.
Rules before build. The chemistry rules were settled in writing before any topic was built, so every topic follows them.
Scoped from the field. I studied how existing worksheet generators work, assessed their depth hands-on, and catalogued topics and features to decide what a teacher-grade chemistry generator needs.
Keys a teacher can defend. Where a reasonable route reaches a neighbouring value, the key names it. Wrong-answer choices stay clear of every accepted answer.
Difficulty design. Options control the number of steps, the style of the given values and whether an equation arrives balanced, unbalanced or as a “balance it first” part. Some problems run in real-world scenes so the numbers carry meaning.
Sheets built for the classroom. Topics come as lists or table-completion sets, in Version A and a shuffled Version B, with name, date and period blanks and an answers view for the key.
An AI coding assistant wrote the software from my written specification. I reviewed the output, and each finding became a rule in the specification, so one fix applies to every sheet that follows.
The specification is where the decisions live: what counts as a correct answer, what makes a problem harder, which conventions a teacher can change. The AI could build quickly because those decisions were already made.
I am actively exploring new instructional design and e-learning development roles. If your team is looking to expand or needs someone to lead complex learning initiatives, reach out at ivanhubert.juan@gmail.com
© 2026 Ivan Hubert C. Juan