RESEARCH INTERESTS

automatic speech recognition, natural language processing, large language model, speech enhancement, machine learning, deep learning, acoustic signal processing

PROFESSIONAL EXPERIENCE 

・SB Intuitions, Nov 2024-present: Researcher 

- Development LLM-based multimodal speech models

・Money Forward Lab, May 2023-Oct 2024: Senior researcher 

- Development to Integrate large language models to commercial products in backoffice and finance domains 

- Released open source large language model as result of joint research with the Institute of Physical and Chemical Research (RIKEN) 

- Research for table question answering task and large language model improvement 

・lilo company, Sep 2023-Dec 2023: Software engineer (side job) 

- Development text-to-speech models for YouTube character voice 

・rinna Co., Ltd., Apr 2022-May 2023: Data scientist 

- Analysis and evaluate performance of large language models for conversation task 

- Development to Integrate large language models to commercial products 

- Research to improve large language models 

・Advanced Media, Inc., Apr 2017-Apr 2022: Software engineer 

- Automatic speech recognition research and development 

- Lead Natural language processing research team for named entity recognition competitions

・NTT Media Intelligence laboratory, Oct 2015: Intern student 

- Natural language processing research 

・Panasonic Corporation, Sep 2013: Intern student

UNIVERSITY DEGREES 

・Hosei University, Japan 

- April 2015-March 2017: MS in Computer Science 

- April 2011-March 2015: BS in Computer Science

AWARDS 

・2022 1st place in NTCIR-16 Real-MedNLP subtask2 (Few-resource Named Entity Recognition) 

・2022 1st place in NTCIR-16 Real-MedNLP subtask1 on radiology report dataset (Few-resource Named Entity Recognition) 

・2017 Student Encouragement Award of IPSJ National Convention (IPSJ) 

OPEN SOURCE SOFTWARE 

・houou: large language model trained by Japanese instruction data created from full scratch (joint research with Riken), 

- https://huggingface.co/moneyforward/houou-instruction-7b-v1

- https://huggingface.co/moneyforward/houou-instruction-7b-v2  

- https://huggingface.co/moneyforward/houou-instruction-7b-v3 

・Beamforming-for-speech-enhancement (My software project on GitHub has received over 270 stars) 

- https://github.com/AkojimaSLP/Beamforming-for-speech-enhancement

-  Used by Alibaba research.

• Referenced in Dianwen Ng, et al, “Small footprint multi-channel network for keyword spotting with centroid based awareness,”  Interspeech, 2024.

MEMBER 

・Information Processing Society of Japan 

・The Acoustical Society of Japan 

・The Association for Natural Language Processing

VOLUNTEER WORKS 

・2024-2025: A Member of Students and Young Researchers Forum of The ASJ, Japan 

・2021-2024: The vice representative of Students and Young Researchers Forum of The ASJ, Japan 

・2019-2021: A Member of Students and Young Researchers Forum of The ASJ, Japan

OTHERS 

・Google Scholar

- https://scholar.google.com/citations?user=y__gbBkAAAAJ&hl=ja 

・Github 

- https://github.com/AkojimaSLP

・Linkedin

• www.linkedin.com/in/atsushi-kojima-417166176