AI Engineer
SIND-AI
Developing Japanese ASR and agentic workflows for clinical documentation.
Phurich Saengthong
I work on AI research and engineering, including anomalous sound detection, representation learning, post-training of audio language models, and agentic workflow engineering.
Experience
Research Intern
Hitachi
Researched post-training and evaluation of audio representations for anomalous sound detection.
Research / Teaching Assistant
Shinozaki Laboratory, Institute of Science Tokyo
Researching anomalous sound detection under domain shift and low-shot conditions.
Graduate Student Researcher
National Electronics and Computer Technology Center
Developed thermal-comfort control systems and human-activity recognition models.
Research Assistant
King Mongkut’s Institute of Technology Ladkrabang
Conducted thermal-comfort experiments, data analysis, and reporting.
Education
Ph.D. in Information and Communications Engineering
Institute of Science Tokyo (formerly Tokyo Institute of Technology)
Toward a Generalized Anomalous Sound Detection System. Advisor: Professor Takahiro Shinozaki.
M.E. in AI/IoT Engineering
TAIST–Tokyo Tech, Sirindhorn International Institute of Technology, Thammasat University
B.E. in Mechanical Engineering
King Mongkut’s Institute of Technology Ladkrabang
Competitions
2nd place · DCASE 2025 Task 2
GenRep
Developed an anomalous sound detection system using large-scale pre-trained audio representations.
7th place · MLC-SLM 2025 Task II
Unified Speech LLM
Developed a multilingual speech language model for joint speaker diarization and speech recognition.
Publications
2026
-
Sub-band spectral matching with localized score aggregation for robust anomalous sound detection.
Phurich Saengthong and T. Shinozaki. Under review at IEEE TASLP.Paper -
Retaining mixture representations for domain generalized anomalous sound detection.
Phurich Saengthong, T. Nishida, K. Dohi, N. Yamashita, and Y. Kawaguchi. EUSIPCO.Paper -
MemNMF: Memory-augmented NMF on LPC spectra for anomalous sound detection.
Phurich Saengthong and T. Shinozaki. EUSIPCO.
2025
-
A unified speech LLM for diarization and speech recognition in multilingual conversations.
Phurich Saengthong, B. Jiaramaneepinit, S. Li, M. Okumura, and T. Shinozaki. MLC-SLM, Interspeech satellite.Paper -
GenRep for first-shot unsupervised anomalous sound detection of DCASE 2025 challenge.
Phurich Saengthong and T. Shinozaki. DCASE technical report.Paper -
Deep generic representations for domain-generalized anomalous sound detection.
Phurich Saengthong and T. Shinozaki. ICASSP.Paper Code
2022
-
Thermal-comfort control using occupancy detection and fuzzy logic for air-conditioning systems.
Phurich Saengthong, C. Charnsripinyo, S. Laitrakun, Y. Hara-Azumi, and S. Deepaisarn. ECTI DAMT & NCON.Paper
2021
-
Energy-efficient airflow control for human thermal comfort using object detection and monocular depth estimation.
Phurich Saengthong, C. Charnsripinyo, and S. Laitrakun. KSE.Paper -
Fusion approaches of heterogeneous multichannel CNN and LSTM models for human activity recognition using wearable sensors.
Phurich Saengthong and S. Laitrakun. ICITEE.Paper
Service
Reviewer for anomalous sound detection and speech:
- ICASSP 2026
- EUSIPCO 2026
- Computer Speech & Language
- IEEE TASLP
Award
NSK Scholarship · NSK Scholarship Foundation, 2022