BrightBook makes it easy for teachers to build lessons that adapted to each student and classroom, bringing personalization into everyday teaching. An award–winning software, it turned lesson planning into a dynamic system that evolved with students as they learn. BrightBook was acquired and merged with DesignMy Education in September 2025.
Project Role
Founded and led BrightBook, overseeing product strategy, design, and development of its AI-powered lesson planning platform. Built the web app using Next.js, Firebase, and Gemini APIs, employing a mixture-of-experts model to generate differentiated instruction aligned with U.S. Common Core, NGSS, and IB standards. Directed the pilot in Boston Public Schools, secured over $20K in non-dilutive funding, and represented the venture in the BU Summer Accelerator and New Venture Competition.
Piloted in the
2nd place out of 160 startups in BU's 2025 New Venture Competition
Panelist at NYU Shanghai
on the future of AI in Education
raised in non–dilutive funding
Stumped gamified student-teacher relationships, creating a school-wide scavenger hunt that built a stronger and more connected community. The successful competition turned faculty members into collectible characters, motivating hundreds of students to forge new bonds outside the classroom.
Project Role
Architected and engineered the software that powered Stumped. A fast and easy-to-use web app, it allowed students to view and guess riddles, redeem their points, and see the overall leaderboard. I built the entire backend and API infrastructure, implementing PII mitigation, FAFSA-compliant data policies, secure authentication, and analytics pipelines for engagement and performance.
students reached
Stumped Cards produced
Awarded Champions of Change by the NYSSBA
NextStepEdu was designed to help students and families apply to college with ease and maximize their financial aid. Like TurboTax for the FAFSA, it guides users step-by-step through every question, adapting to their unique circumstances and turning confusing tax data into clear, personalized answers. NextStep was acquired in late 2025.
Project Role
Engineered a secure end-to-end system integrating Gemini 2.5 Pro and Google Document AI for high-accuracy data extraction and validation. Implemented rigorous PII controls, including encryption at rest and in transit, access-scoped data handling, and anonymization of stored personal identifiers. Embedded an integrated feedback mechanism throughout the interface to facilitate natural, real-time input from beta testers during testing and refinement.
accuracy in data extraction
and document classification*
Tech Stack
*Approx. 96.35% accuracy in extraction and classification across 124 anonymized 1040 and W-2 datasets using Gemini-2.5 Pro + Document AI (n = 124 forms, ±1.3 std)