GPU clusters bottlenecked Kakao's AI. Switching to Cloud TPUs, the team upcycled a 2.1B model into a 13.4B parameter MoE system.
South Korea's leading technology company operates the country's dominant messaging platform along with a vast ecosystem spanning digital payments, mobility, and entertainment.
Developing a proprietary Korean-language AI model required massive computing power, but scaling GPU clusters beyond single server units caused...
“현재 카카오의 대규모 언어 모델은 모두 TPU Trillium으로 학습되고 있습니다. GPU는 익숙하고, 다양한 접근이 가능하긴 하지만 필요한 만큼의 성능을 확보하는 데에 근본적인 어려움이 이어졌습니다. 구글 클라우드의 TPU는 언제든 필요한 만큼 자원을 할당받을 수 있었고 성능이나 모델의 확장도 손쉬웠습니다.”
Mobile messaging platform and internet services for the South Korean market.
Supported Kakao's AI and cloud transition to develop its 'Kanana' language model.
Cloud computing services, AI infrastructure, and data analytics platforms for enterprises.
Kakao's AI model training is part of this use case:
Related implementations across industries and use cases
Engineers spent 90% of time on data prep. New pipelines flipped that to 90% modeling and cut tuning from 7 days to 1 hour.
Data silos capped counselor reach. Now, orchestrated AI agents analyze sleep, nutrition, and training to multiply human expert capacity.
Processing trillion-token datasets took months. A native vector engine cut deduplication costs 5x and doubled processing speed.
Engineers spent 90% of time on data prep. New pipelines flipped that to 90% modeling and cut tuning from 7 days to 1 hour.
Processing 2T tokens on CPUs took days. GPU acceleration cut prep to hours, unlocking a 5% accuracy gain.
Scattered data and basic coding tools bottlenecked engineers. A 9-agent AI workflow shifts them from writing code to directing AI teams.
Sequential AI testing bottlenecked development. Engineers built a concurrent, code-first pipeline to evaluate agent responses in seconds.
On-premise systems, dispersed and brittle, bottlenecked every release. AI agents now run routine dev steps — hours cut to minutes.
A mistranslated word could derail global R&D projects. Now, researchers instantly refine technical papers & communicate seamlessly across languages.
GPU clusters bottlenecked Kakao's AI. Switching to Cloud TPUs, the team upcycled a 2.1B model into a 13.4B parameter MoE system.
South Korea's leading technology company operates the country's dominant messaging platform along with a vast ecosystem spanning digital payments, mobility, and entertainment.
Developing a proprietary Korean-language AI model required massive computing power, but scaling GPU clusters beyond single server units caused...
“현재 카카오의 대규모 언어 모델은 모두 TPU Trillium으로 학습되고 있습니다. GPU는 익숙하고, 다양한 접근이 가능하긴 하지만 필요한 만큼의 성능을 확보하는 데에 근본적인 어려움이 이어졌습니다. 구글 클라우드의 TPU는 언제든 필요한 만큼 자원을 할당받을 수 있었고 성능이나 모델의 확장도 손쉬웠습니다.”
Mobile messaging platform and internet services for the South Korean market.
Supported Kakao's AI and cloud transition to develop its 'Kanana' language model.
Cloud computing services, AI infrastructure, and data analytics platforms for enterprises.
Kakao's AI model training is part of this use case:
Related implementations across industries and use cases
Engineers spent 90% of time on data prep. New pipelines flipped that to 90% modeling and cut tuning from 7 days to 1 hour.
Data silos capped counselor reach. Now, orchestrated AI agents analyze sleep, nutrition, and training to multiply human expert capacity.
Processing trillion-token datasets took months. A native vector engine cut deduplication costs 5x and doubled processing speed.
Engineers spent 90% of time on data prep. New pipelines flipped that to 90% modeling and cut tuning from 7 days to 1 hour.
Processing 2T tokens on CPUs took days. GPU acceleration cut prep to hours, unlocking a 5% accuracy gain.
Scattered data and basic coding tools bottlenecked engineers. A 9-agent AI workflow shifts them from writing code to directing AI teams.
Sequential AI testing bottlenecked development. Engineers built a concurrent, code-first pipeline to evaluate agent responses in seconds.
On-premise systems, dispersed and brittle, bottlenecked every release. AI agents now run routine dev steps — hours cut to minutes.
A mistranslated word could derail global R&D projects. Now, researchers instantly refine technical papers & communicate seamlessly across languages.