About
Jin Heo is a Senior Researcher at Dolby Laboratories and holds a Ph.D. in Computer Science from Georgia Institute of Technology, where he conducted research in ML inference serving, adaptive resource management, and distributed systems under Prof. Ada Gavrilovska.
His research focuses on efficiently serving real-time workloads under resource and latency constraints, with applications spanning XR, cloud gaming, and real-time perception.
At Dolby, he extends this work to production-scale systems, architecting ML-enabled cloud and streaming infrastructure from research models to distributed service deployment.
Technical Expertise & Focus
- ML Inference Serving & Adaptive Resource Management:
Designing real-time inference serving and scheduling systems that adapt resource allocation to workload dynamics, result effectiveness, and resource constraints.
- Real-Time & Distributed Systems:
Building distributed systems for latency-sensitive workloads across resource-constrained edge and scalable cloud environments, with applications in XR, cloud gaming, and streaming perception.
- ML & Cloud Infrastructure:
Architecting end-to-end ML-enabled cloud systems, including distributed microservices, model deployment, infrastructure orchestration, and high-throughput streaming data pipelines.
Background
His interest in systems and immersive computing grew from an early career as a professional gamer, which sparked his curiosity about the intersection of computer science, games, and interactive experiences.
This early interest ultimately led to research on high-performance systems for XR, cloud gaming, and real-time perception.
์๊ฐ
ํ์ง (Jin Heo) ๋ ํ์ฌ Dolby Laboratories์ Senior Researcher๋ก ์ฌ์ง ์ค์ด๋ฉฐ, Georgia Institute of Technology์์ ์ปดํจํฐ๊ณผํ ๋ฐ์ฌํ์๋ฅผ ๋ฐ์์ต๋๋ค.
Georgia Tech์์๋ Ada Gavrilovska ๊ต์์ ์ง๋ ์๋ ML inference serving, adaptive resource management, distributed systems๋ฅผ ์ฐ๊ตฌํ์ต๋๋ค.
์ฃผ์ ์ฐ๊ตฌ๋ ์์ ๋ฐ ์ง์ฐ์๊ฐ ์ ์ฝ์ด ์๋ ํ๊ฒฝ์์ ์ค์๊ฐ ์ํฌ๋ก๋๋ฅผ ํจ์จ์ ์ผ๋ก ์ฒ๋ฆฌํ๊ณ ์๋นํ๋ ์์คํ
์ ์ด์ ์ ๋๋ฉฐ, XR, ํด๋ผ์ฐ๋ ๊ฒ์ด๋ฐ, ์ค์๊ฐ perception ๋ฑ์ ์์ฉ ๋ถ์ผ๋ฅผ ๋ค๋ฃน๋๋ค.
Dolby์์๋ ์ด๋ฌํ ์ฐ๊ตฌ ๊ฒฝํ์ production-scale ์์คํ
์ผ๋ก ํ์ฅํ์ฌ, ์ฐ๊ตฌ ๋ชจ๋ธ๋ถํฐ ๋ถ์ฐ ์๋น์ค ๋ฐฐํฌ๊น์ง ์์ฐ๋ฅด๋ ML ๊ธฐ๋ฐ ํด๋ผ์ฐ๋ ๋ฐ ์คํธ๋ฆฌ๋ฐ ์ธํ๋ผ๋ฅผ ์ค๊ณํ๊ณ ์์ต๋๋ค.
์ฃผ์ ์ฐ๊ตฌ ๋ถ์ผ ๋ฐ ๊ธฐ์
- ML Inference Serving & Adaptive Resource Management:
์ํฌ๋ก๋์ ๋ณํ, inference ๊ฒฐ๊ณผ์ effectiveness, ๊ทธ๋ฆฌ๊ณ ์์คํ
์ ์์ ์ ์ฝ์ ๋ฐ๋ผ ์์ ํ ๋น์ ๋์ ์ผ๋ก ์กฐ์ ํ๋ ์ค์๊ฐ inference serving ๋ฐ scheduling ์์คํ
์ ์ค๊ณํฉ๋๋ค.
- Real-Time & Distributed Systems:
์ง์ฐ์๊ฐ์ ๋ฏผ๊ฐํ ์ํฌ๋ก๋๋ฅผ resource-constrained edge ํ๊ฒฝ๋ถํฐ scalable cloud ํ๊ฒฝ๊น์ง ํจ์จ์ ์ผ๋ก ์ฒ๋ฆฌํ๊ธฐ ์ํ ๋ถ์ฐ ์์คํ
์ ๊ตฌ์ถํ๋ฉฐ, XR, ํด๋ผ์ฐ๋ ๊ฒ์ด๋ฐ, streaming perception ๋ฑ์ ์์ฉ ๋ถ์ผ๋ฅผ ๋ค๋ฃน๋๋ค.
- ML & Cloud Infrastructure:
๋ถ์ฐ ๋ง์ดํฌ๋ก์๋น์ค, ๋ชจ๋ธ ๋ฐฐํฌ, ์ธํ๋ผ ์ค์ผ์คํธ๋ ์ด์
, ๋๊ท๋ชจ ์คํธ๋ฆฌ๋ฐ ๋ฐ์ดํฐ ํ์ดํ๋ผ์ธ ๋ฑ์ ํฌํจํ๋ end-to-end ML ๊ธฐ๋ฐ ํด๋ผ์ฐ๋ ์์คํ
์ ์ค๊ณํฉ๋๋ค.
๋ฐฐ๊ฒฝ
ํ๋ก๊ฒ์ด๋จธ๋ก์์ ์ด๊ธฐ ๊ฒฝํ์ ๊ณ๊ธฐ๋ก ์์คํ
๊ณผ immersive computing์ ๊ด์ฌ์ ๊ฐ๊ฒ ๋์์ต๋๋ค.
๋น์ ์ปดํจํฐ๊ณผํ๊ณผ ๊ฒ์ ๋ฐ ์ธํฐ๋ํฐ๋ธ ๊ฒฝํ์ ์ ์ ์ ๊ด์ฌ์ ๊ฐ์ง๊ฒ ๋์๊ณ , ์ด๋ฌํ ๊ด์ฌ์ ์ดํ XR, ํด๋ผ์ฐ๋ ๊ฒ์ด๋ฐ, ์ค์๊ฐ perception์ ์ํ ๊ณ ์ฑ๋ฅ ์์คํ
์ฐ๊ตฌ๋ก ์ด์ด์ก์ต๋๋ค.