Hongbee Park

At Scale AI, I specialize in human-in-the-loop evaluation for generative AI models like GPT, Gemini, and Grok. I investigate how each model interprets prompts and iteratively refines its chain-of-thought reasoning. By intentionally testing edge cases and failure modes, I shape model behavior and inform improvements across product iterations.

This role has deepened my backend understanding of how large language models operate and positioned me at the intersection of product growth, research, and AI alignment.

Over four years at Toss, Korea's leading fintech platform, I built and scaled growth systems across the full user lifecycle — from acquisition and activation to conversion and long-term retention.

As Growth Domain Lead, I orchestrated large-scale experiments across product, engineering, design, data, and policy teams, translating behavioral insights into measurable growth levers. Through funnel redesign, activation optimization, and platform-level experimentation, I helped drive over 10M MAU growth in a highly regulated, hyper-competitive market.

Beyond individual feature launches, my focus was on designing repeatable growth engines — building the tooling, metrics, and operating rhythms that enabled fast iteration, informed decision-making, and sustainable product-market fit at scale.

At the Korean Digital Platform Government, I worked on national-scale digital infrastructure projects responsible for serving the entire population during the COVID crisis.

Under extreme time pressure and policy constraints, I helped design and launch mission-critical systems that digitized citizen identity, public records, and financial workflows — ensuring stability, security, and public trust for products used by over 40 million people.

My work included architecting API-based tax notification systems, enabling end-to-end electronic delivery of government communications, and contributing to the national MyData initiative — a platform that standardized and digitized citizen data across ministries. These systems replaced manual, fragmented processes with interoperable, real-time infrastructure that could operate reliably at country-wide scale.

Through this experience, I learned how to design technology not just for usability, but for governance — balancing speed, reliability, policy, and public accountability in high-stakes environments.

At LG, I led cross-channel product development in e-commerce and payment systems, improving checkout conversion by 5% and driving $1M+/month in additional sales.

I also bridged offline SCM with real-time online tools, designed backend payment architecture, and worked closely with dev teams on system prototyping and deployment. These years taught me how to think in systems, quantify success, and balance business and engineering trade-offs.

AIRQUA is a disaster-relief water generator designed for low-literacy users. I led 50+ user interviews and 30+ hardware iterations to design a scalable, intuitive solution.

The project won Harvard's highest innovation honor—personally awarded by President Alan Garber. This experience reinforced my belief that the most impactful technology isn't the most complex, but the most accessible.