Applied AI,
from problem to production.

Applied AI engineer and Seoul National University Ph.D. building deployable AI systems across enterprise, edge, and consumer settings.

View experience ↘View selected work ↗

Profile

I turn ambiguous real-world problems into AI systems that work under real product and deployment constraints.

My work spans customer discovery, technical scoping, architecture, hands-on implementation, evaluation, field deployment, and iteration. I choose models and tools based on accuracy, latency, memory, privacy, device cost, and measurable product value.

9+years of applied AI
7person R&D team led
12/42patents registered / filed
2/5+apps shipped / in development

Experience

FEB. 2025 — NOW

Samsung Electronics

AI Engineer · On-device Speech Processing

  • Develop and optimize on-device speech systems, including Overlapped Speech Detection and Speech Separation, with deployment and evaluation on resource-constrained devices.
  • Lead an ultra-lightweight VAD project for Galaxy Buds with Samsung R&D Institute Bangalore, translating product constraints into failure scenarios, inference-resource targets, technical requirements, evaluation criteria, and delivery scope.
  • Built a web-based evaluation framework and automated experiment pipeline that replaced repetitive manual testing, reduced operator-dependent error, and accelerated performance evaluation, failure analysis, and model iteration.

FEB. 2026 — NOW

Across Realities ↗

Founder & Independent Developer

  • Own end-to-end product development from problem definition and architecture through AI/ML design, implementation, UI/UX, testing, and release; launched two consumer apps with additional products in development.
  • Prototyped an AI safety system for elderly-care environments, translating per-room hardware cost and heterogeneous-device requirements into model memory and latency constraints.
  • Evaluated multiple on-device LLM configurations for contextual emergency classification, including context-length reduction and local inference trade-offs; tested local and cloud inference with Ollama on Mac hardware and GPU instances, then shifted toward a lighter-weight architecture when LLM compute and latency costs outweighed expected product benefit.

2017 — 2024

Cochl

Co-founder & Lead Research Scientist

Built B2B AI products from customer discovery through field deployment.

  • Led a 7-person R&D team and owned the B2B AI product lifecycle from client discovery and use-case analysis through architecture, hands-on implementation, field testing, deployment, and iteration.
  • Worked directly with customers to translate business needs and operational constraints into technical proposals, evaluation plans, and deployment strategies.
  • Optimized edge sound-recognition systems using pruning, fine-tuning, INT8 quantization, and ARM CPU optimization; reduced false detections by 83% while increasing recall by 22%.
  • Delivered device-integrated B2B AI solutions for nursing-home safety and factory anomaly detection; secured 12 registered international patents and presented 7 papers at conferences including ICASSP and DCASE.

JAN. — FEB. 2014

Samsung Medison

Research Intern

Developed an automatic liver-cirrhosis diagnosis system leveraging ultrasound imaging.

Selected Product Work

Across Realities ↗

Here, I Hear

AI safety system for elderly-care environments — in development

→

Locale Sight

View product site

↗

Memoir Log

View product site

↗

Applied AI Focus

◌

Problem-first AI systems

Translate customer, product, and operational constraints into architecture, evaluation plans, implementation choices, and deployment decisions.

⌁

Efficient on-device AI

Speech and audio models, VAD, pruning, fine-tuning, INT8 quantization, ARM optimization, and latency/memory trade-offs for resource-constrained devices.

⌬

LLM systems & foundations

Local and cloud LLM inference, Ollama, context-length versus memory trade-offs, and hands-on implementation and training of a small decoder-only Transformer.

Selected Applied Work & Demos

120+ devices prepared for a real site

Configured and deployed 120+ Linux devices for a Japanese nursing home demonstration.

Education

2012 — 2018

Ph.D. in Engineering, Seoul National University

  • Music and Audio Research Group (MARG), Graduate School of Convergence Science and Technology.
  • Topic: Machine-learning approaches to music source separation and preprocessing of audio signals.
  • Thesis: Unsupervised approach to music source separation using generalized Dirichlet prior. Advisor: Kyogu Lee.

2010 — 2012

M.S. in Electrical Engineering, Seoul National University

  • Transceiver Technology Laboratory, Department of Electrical Engineering.
  • Topic: Physical-layer wireless communication, including LTE-Advanced, MIMO, and opportunistic channel-information feedback.
  • Thesis: Opportunistic channel state information feedback in multi-user MIMO environments. Advisor: Yong-Hwan Lee.

2006 — 2010

B.S. in Electrical Engineering, Seoul National University

  • Department of Electrical Engineering, College of Engineering.
  • Thesis: Face detection using Haar features. Advisor: Jong-ho Choi.

Publications & Patents

Papers15
  1. Park, Jeongsoo, Dong-Gyun Han, Hyoung Sul La, Sangmin Lee, Yoonchang Han, and Eun-Jin Yang. “Cross-domain Sound Recognition for Efficient Underwater Data Analysis.” APSIPA ASC, 2023.
  2. Jeong, Il-Young, and Jeongsoo Park. “CochlScene: Acquisition of acoustic scene data using crowdsourcing.” APSIPA ASC, 2022.
  3. Yang, Jinsung, Jeongsoo Park, Melanie Koehler, Joshua Simpson, Daniel Luque, Javier M. Rodríguez, and David Alsteens. “Rotavirus Binding to Cell Surface Receptors Directly Recruiting α2 Integrin.” Advanced NanoBiomed Research 1.12, 2021.
  4. Chang, Sungkyun, Donmoon Lee, Jeongsoo Park, Hyungui Lim, Kyogu Lee, Karam Ko, and Yoonchang Han. “Neural audio fingerprint for high-specific audio retrieval based on contrastive learning.” ICASSP, 2021.
  5. Lee, Donmoon, Jaejun Lee, Jeongsoo Park, and Kyogu Lee. “Enhancing music features by knowledge transfer from user-item log data.” ICASSP, 2019.
  6. Park, Jeongsoo, Jaeyoung Shin, and Kyogu Lee. “Separation of instrument sounds using non-negative matrix factorization with spectral envelope constraints.” arXiv:1801.04081, 2018.
  7. Lim, Hyungui, Jeong-Soo Park, and Yoonchang Han. “Rare Sound Event Detection Using 1D Convolutional Recurrent Neural Networks.” DCASE, 2017.
  8. Han, Yoonchang, Jeongsoo Park, and Kyogu Lee. “Convolutional Neural Networks with Binaural Representations and Background Subtraction for Acoustic Scene Classification.” DCASE, 2017.
  9. Park, Jeongsoo, Jaeyoung Shin, and Kyogu Lee. “Exploiting continuity/discontinuity of basis vectors in spectrogram decomposition for harmonic-percussive sound separation.” IEEE/ACM TASLP 25.5, 2017.
  10. Park, Jeongsoo, and Kyogu Lee. “Harmonic-Percussive Source Separation Using Harmonicity and Sparsity Constraints.” ISMIR, 2015.
  11. Oh, Changhoon, Jeongsoo Park, and Bongwon Suh. “Gravity: automatic location tracking system between a car and a pedestrian.” MobileHCI, 2014.
  12. Park, Jeongsoo, and Kyogu Lee. “Separation of monophonic music signal based on user-guided onset information.” International Congress on Sound and Vibration, 2014.
  13. Park, Jeong-Soo, Yong-Suk Byun, Moon-Hyung Yoon, and Yong-Hwan Lee. “Opportunistic feedback of channel information in multi-user MIMO environments.” IEEE WCNC, 2013.
  14. Heo, Gil-Su, Jeong-Soo Park, Yong-Suk Byun, and Yong-Hwan Lee. “Opportunistic user scheduling with normalized CQI in heterogeneous channel environments.” IEEE WoWMoM, 2012.
  15. Park, Jeong-Soo, Gil-Su Heo, and Yong-Hwan Lee. “Capacity maximizing cell-selection in heterogeneous cellular networks.” SoftCOM, 2011.
Selected patent records20
  1. Method, device, and program for improving accuracy of sound data recognition
  2. Method, device, and program for providing matching information through analysis of sound information
  3. Audio quality conversion device and control method therefor
  4. Audio signal-based device and control method thereof
  5. Lifelog device utilizing audio recognition, and method therefor
  6. Active sound device utilizing audio recognition
  7. Device for detecting music data from video contents, and method for controlling same
  8. Augmented reality device performing audio recognition and control method therefor
  9. Audio data identification apparatus
  10. Audio tagging system using densely connected convolutional networks
  11. Image processing device and control method therefor
  12. Control apparatus for electronic device using non-vocal audio information and control method thereof
  13. Audio information collecting apparatus and control method thereof
  14. Sleep apnea severity testing device
  15. Method and apparatus for processing audio signal
  16. Boosting method and apparatus for harmonic components of audio signals
  17. Method of harmonic percussive source separation using harmonicity and sparsity constraints
  18. Method and apparatus for selecting cell in heterogeneous cell wireless communication system
  19. Method for cooperative signal transmitting in multi-cell cooperative communication system
  20. A method of cooperative transmission in multi-user multi-antenna wireless communication systems

Professional Activities

Reviewer Experience

IEEE/ACM TASLP, PLOS ONE, AES Semantic Audio, DAFx, ISMIR, ICASSP, and IEEE MMSP.

Teaching Experience

Introduction to Machine Learning, a Big Data Engineer course for 100+ Seoul citizens hosted by Seoul National University’s Big Data Institute (Jul.–Sept. 2017).

Cross-disciplinary Collaboration

Automated information extraction from biological experiment data using Python tools, contributing to published interdisciplinary research.

View paper ↗

International Industry Engagement

Represented Cochl at networking events in San Jose (May–Sept. 2022), translating technical capabilities into product and deployment conversations with industry stakeholders.

Skills & Technologies

Core Expertise

Applied AI SystemsEnd-to-End Product LifecycleEvaluationOn-device AIEdge ComputingAudio & Speech AI

AI & Machine Learning

LLM SystemsDecoder-only TransformersTensorFlowFine-TuningModel PruningINT8 Quantization

Languages & Frameworks

PythonC++KotlinJavaJavaScript

Tools & Infrastructure

OllamaGitGerritDockerUbuntuAndroid StudioAWSGoogle Cloud Platform

Languages

EnglishFull Professional Proficiency
KoreanNative

Contact

Professional inquiries and collaboration.

psprink@snu.ac.kr ↗
LinkedIn ↗Google Scholar ↗Homepage ↗