- Built the core scoring engine, the product's primary differentiator: applied parallel algorithms for batch score computation and an incremental update strategy (subtract the old contribution, add the new — avoiding full recompute on every edit), cutting CPU usage ~90% on score updates.
- Designed the architecture and PostgreSQL schema end-to-end: modeled the multi-tenant org/employee/competency hierarchy across 79 tables with tenant isolation enforced at the service layer (rather than row-level) to keep schema and query paths simple.
- Led an 8-engineer team from MVP to production: ran code reviews, drove technical direction, and mentored junior devs through MVP shipping and post-launch iteration.
- Drove the technology selection for the new product: evaluated stacks across hiring availability, ecosystem maturity, and TypeScript-level type safety; landed on Node.js + TypeScript + Prisma + PostgreSQL.
- Built a Puppeteer + HTML-templates PDF report generator producing 2K+ branded reports/month across customer orgs, wired into transactional email and replacing manual reporting workflows.
- Adopted PgBoss (PostgreSQL-backed job queue) over Redis-based alternatives to avoid adding a separate infra dependency and preserve transactional consistency with the main DB; offloaded transactional email to async workers, decoupling outbound delivery from API response latency.