She taught machines to read Korean. Now she's teaching them to work.
In 2014, a PhD student at Seoul National University released a small Python package into the world. She called it KoNLPy. The idea was simple: give Korean developers the same natural language processing tools that English speakers had been taking for granted for years. No institution paid her to do it. No company mandated it. She just saw the gap and closed it.
That instinct - spot what's missing, build it, share it freely - has defined every chapter of Lucy Park's career. The handle she chose for herself, "echojuliett," is NATO alphabet for E and J, her initials. Precise, efficient, no decoration. Very on brand.
KoNLPy became the Korean NLP library. Not one of several options. The one universities teach from, the one companies ship with, the one Stack Overflow answers reference. Park didn't market it. She just made it work well and put it where people could find it.
"Technology is not the goal but an enabler."
After finishing her PhD in Data Mining at Seoul National University in 2016, she joined NAVER - South Korea's Google equivalent - to lead the Papago machine translation team. Papago is the translation app Koreans actually use. Building it meant confronting a challenge that makes English-to-French translation look trivial: Korean honorifics. The language has grammatically distinct levels of formality baked into every sentence, and getting them wrong doesn't just sound awkward - it signals something about who you think the other person is. Park built a system that could navigate that. It launched in 2017 and became widely used among Korean language learners worldwide.
Four years at NAVER, then a pivot. In October 2020, Park co-founded Upstage with CEO Sung-hoon Kim and CTO Stan Lee. They shared a conviction: AI had become extraordinary in the lab and frustrating in production. The gap between what models could do in benchmarks and what enterprises could actually deploy reliably was enormous. Upstage was built to close that gap.
Park was initially Chief Scientific Officer at Upstage, steering the research agenda. She transitioned to Chief Product Officer in June 2024 - her own description of the shift was "excited for this new chapter and hopeful that it will lead to successful productization of Upstage's services." Understated. The company hit unicorn status less than two years later.
The product Park built her CPO chapter around is twofold: Solar, Upstage's proprietary large language model family, and Document Parse, their enterprise document intelligence platform. The combination targets industries that run on paper - insurance claims, legal filings, financial reports, medical records - and promises something remarkable for regulated industries: accuracy you can explain, compliance you can prove, deployment options that don't require handing your data to a cloud provider.
The numbers bear it out. Upstage is automating more than 60% of insurance claims processing in Korea. That's not a proof of concept - that's operational infrastructure at national scale. When insurers trust your AI with claims, they're trusting it with money and with customers. The bar is different from a chatbot that gives interesting answers.
In 2024, Park took on a third title: US CEO. Upstage opened its North American headquarters in San Jose, California, with Park leading the charge from the Bay Area. Speaking at Qualcomm's Snapdragon Summit in November 2024, she presented Solar Box - an on-device AI solution that runs Solar-class models on Snapdragon hardware. The direction is deliberate: enterprise customers who can't put sensitive documents in the cloud get a path to powerful AI without the compliance risk.
In April 2026, Upstage completed its Series C at $120 million, pushing total funding past $326 million and tipping the valuation over $1 billion. The company became the first South Korean generative AI startup to reach unicorn status. The investors include Hyundai Motor, Kia, SK Networks, and Sagemaker Partners - an unusual coalition that reflects Upstage's positioning: not purely a tech play but an industrial AI company with genuine enterprise traction.
Park's trajectory from open-source contributor to unicorn founder is not a pivot story. The thread is continuous. KoNLPy lowered barriers to Korean NLP. Papago lowered barriers to Korean language access. KLUE - the Korean Language Understanding Evaluation benchmark she co-led at Upstage - lowered barriers to rigorous Korean AI research. Solar and Document Parse are lowering barriers to enterprise AI deployment in regulated industries. Same mission, bigger scope, higher stakes.
