Doh-Suk Kim

Redmond, WA, United States0

Applied Research & Technology Leader | Audio, Speech & Voice AI | Perceptual Science, Evaluation & Systems | ex-Dolby, Lucent, Samsung | Startup Advisor

I am an applied research and technology leader with 28+ years of experience at the intersection of human perception, audio and speech processing, communications, and AI. My work has focused on a fundamental challenge: how do we measure complex audio and communication systems in ways that reflect human experience - and translate those measurements into better technologies, products, and engineering decisions. Throughout my career at Microsoft, Dolby, Typecast, Huawei, Bell Labs/Alcatel-Lucent, and Samsung, I have developed perceptually grounded evaluation methodologies, objective quality models, benchmarking frameworks, and advanced audio and communications technologies. My expertise spans psychoacoustics, speech/audio quality, real-time communications, spatial audio, speech enhancement, echo suppression, QoE, and ML-based perceptual modeling. ___Current focus - Audio Systems and Voice AI Evaluation:___ At Microsoft, I am developing automated audio-quality evaluation and root-cause-analysis frameworks for Windows and emerging device experiences. My work connects system-level measurements and perceptual KPIs to engineering decisions and hardware/OEM roadmaps. I am also exploring how upstream acoustic and speech-processing degradations propagate into downstream LLM-based voice-agent responses and overall interaction quality. ___Research and Technical Leadership:___ I have spent more than 15 years leading multidisciplinary research and engineering initiatives, including research teams, technology roadmaps, perceptual benchmarking, and the translation of advanced research into practical products and platforms. My contributions include: - 60+ publications and standards contributions in audio, speech, perception, and communications. - 20+ patents spanning speech/audio quality, perceptual processing, and communications technologies. - Significant contributions to ITU-T and ANSI standards, including non-intrusive speech-quality assessment and multi-party conferencing evaluation. - Leadership in subjective and objective evaluation, perceptual modeling, and technology benchmarking. ___Looking Ahead___ I am particularly interested in the convergence of perceptual science, real-time voice systems, and multimodal AI - where measuring human experience and downstream task performance can play a central role in building more reliable, natural, and effective AI-powered systems.

AI Fluency

No AI stories yet

Nothing posted yet.

Experience

Audio

Microsoft

Jun 2025 — Present

Developing advanced audio-system evaluation methodologies and automated analysis frameworks for Windows and emerging AI-powered device experiences. - Developed an automated Windows audio-quality evaluation framework integrating scenario-driven testing, objective and perceptual metrics, and system-level diagnostics. - Developed automated root-cause-analysis capabilities for Project Solara agent devices, reducing evaluation failure rates by more than 10×. - Investigate end-to-end voice AI quality, including how upstream acoustic and speech-processing degradations propagate through downstream LLM-based agent responses. - Collaborate across engineering, hardware, and product teams to improve the reliability, scalability, and diagnostic value of audio-system evaluation.

Career Break

Unknown

Apr 2025 — Jun 2025

Travelling

Advisory Consultant | VP of Engineering

Typecast

Feb 2024 — Mar 2025 · San Francisco Bay Area

- Spearheaded the AI research team to design and manage a data pipeline for training advanced voice generation AI models, ensuring efficiency and scalability. - Led the engineering team in developing a web-based platform for remote subjective testing, enabling efficient AI model performance evaluation. - Led the development of subjective and objective evaluation pipelines for AI voice models, achieving 50% cost reduction and 80% improvement in the audio performance assurance process timeline.

Director of Advanced Media Systems | Director of Advanced Performance Measurements

Dolby Laboratories

Jan 2012 — Oct 2023 · Sunnyvale, CA

Dolby Voice spatial audio-video conferencing & Perception-based personalized video streaming - Led defensibility strategy and cross-organizational projects to benchmark Dolby Voice against competitive audio conferencing devices and systems, resulting in the development of systematic performance management framework and white papers demonstrating quantitatively articulated competitive marketing advantages. - Managed and mentored a team of researchers, overseeing project resourcing and career development. - Drove development of subjective and objective methods for measuring the performance of multi-party and spatial audio-video conferencing, resulting in the adoption of 3 new standard recommendations ITU-T Rec. P.340 Annex B (objective method), ITU-T Rec. P.1311 and P.1312 (subjective methods). - Drove the development of in-service live conference performance measurement models tailored to perceptual KPI’s, enabling efficient monitoring of commercial Dolby Voice live traffic for multiple years. - Led research in personalized adaptive media streaming optimization, achieving 30-70% bandwidth savings through innovative sensorless and sensor-based user space understanding.

Credentials

Publication

Hybrid Images for Personalized Media Streaming Optimization

Efficiently transmitting media while maintaining or improving the quality of experience (QoE) poses a constant challenge. On the media playback side, there exists a variety of factors affecting the resulting QoE perceived by end users, such as device characteristics, viewing conditions, and individual visual acuity. Understanding the characteristics of playback-side context can provide opportunities for optimizing media delivery tailored to each user. — We present a novel approach to estimating viewers’ visible frequency ranges by creating hybrid images in a quantitatively controlled manner. The proposed approach, which can be applied to adaptive bitrate (ABR) streaming optimization, combines low- and high-frequency components of two different images providing multiscale image perception. The proposed method is designed to adjust the hybrid image size and the cutoff frequencies of lowpass and highpass filters of hybrid image synthesis according to the given frequency range under test. This method can be used to estimate the minimum video resolution for maintaining maximum QoE, in any viewing conditions and devices. — Our experimental results reveal a strong correlation between the minimum resolution obtained from hybrid image perception tasks and the traditional just noticeable difference (JND) of real images of different resolutions when presented side-by-side. These JND tasks exhibit a wider spread both in the measured minimum resolution and elapsed time to complete the tasks, potentially due to the complexity of the cognitive process and the variability caused by contents and subjects. In contrast, the hybrid image technique achieves more stable results with much less effort from subjects.

Publication

Objective test method for measuring the performance of conference phones using real speech signals in multiple-talker scenarios - An experiment in reverberant conditions

Publication

A proposal for objective method to measure the performance of conference phones using real speech signals for multiple talker scenarios

Publication

Objective performance measurement method for multi-talker scenarios using real speech signals