Yureka — AI Study Platform
An agentic study planner that turns messy student learning data into personalized guidance.
Founding Engineer & Product Manager · Yureka (AI Education Startup)
The problem
Students and mentors sit on scattered learning data — quiz scores, session notes, study logs — but nobody has time to turn it into a concrete plan. A tutoring business can't scale personalized guidance when it depends on a human reading every student's history by hand.
Goal & constraints
Ship a production platform, as the sole engineer, that ingests real student learning data and produces personalized study plans and mentor insights automatically — reliable enough to charge for and operable by a small team.
Key decisions
Build the planner as an agentic pipeline rather than one big prompt.
A single one-shot prompt couldn't reason over messy, multi-source student history reliably. Splitting analysis → planning → recommendation into discrete agent steps made each stage testable and the output auditable when a plan looked wrong.
Sync knowledge from Google Drive instead of building a separate CMS.
Mentors already worked in Google Drive. Meeting them there removed an adoption barrier and let AI agents retrieve and index existing material instead of asking anyone to migrate.
Own the full stack solo on Next.js + AWS.
As the only engineer, I optimized for a stack I could design, deploy, and operate end-to-end without handoffs. At this stage fast iteration mattered more than specialization.
How I built it
Agentic study planner
Built a pipeline that analyzes a student's learning data, generates a personalized study plan, and surfaces mentor-facing insights and actionable recommendations.
Knowledge agents
Developed AI agents for Google Drive knowledge synchronization, retrieval, and automated dashboard generation, tightening the team's internal knowledge workflows.
Cross-border delivery
Led a cross-border engineering collaboration with Singapore Management University — translating product requirements into engineering tasks, coordinating implementation, and reviewing deliverables.
Product & pricing
Defined the roadmap and pricing across planner subscriptions, 1:1 mentoring, and B2B school programs.
Outcome
Shipped the company's first production AI education platform end-to-end and defined the model that landed Yureka's first paid school partnership.
What I took away
- Wearing both PM and engineer hats forced every technical decision to earn its keep against a real pricing and adoption goal.
- Agent pipelines are worth the extra plumbing the moment outputs need to be trusted and debugged, not just demoed.