Thodsaporn Chay-intr, D. Eng.

Head of AI at iApp Technology, AI/ML Engineer, and NLP Specialist

Resume

Click here to download the PDF version of my resume.

Work Experience

  • AI/ML Engineer and Head of AI
    iApp Technology Co., Ltd., Thailand, Jan 2024 - Jan 2025
    • Led AI projects (text, vision, and audio), transforming practical research into production and doubling team efficiency.
    • Managed AI server infrastructure with containers and orchestration, ensuring scalable and reliable ML performance.
    • Developed efficient LLM-based agents with RAG and TensorRT-LLM, achieving 97.67% QA accuracy, 87.53% recall for Thai.
    • Analyzed linguistic aspects of international languages (Thai, Chinese, Japanese) to deliver NLP projects from concept to release.
  • ML Researcher and Lecturer
    Artificial Intelligence Association of Thailand, Thailand, Jan 2018 - Aug 2023
    • Advised professionals and scholars on ML, contributing to the planning and publication of over 20 research papers.
    • Delivered courses on ML/NLP, focusing on algorithms, techniques, and tools (e.g., SVM, parsing, and scikit-learn).
  • Research Assistant
    Tokyo Institute of Technology, Japan, Sep 2019 - Mar 2020
    • Collaborated with a multidisciplinary team to develop NLU/NLG modules for Japanese conversational dialogues.
    • Built a BiLSTM Seq2Seq model with cross-attention in PyTorch for natural text generation, validated via human evaluation.
  • ML Engineer and Researcher
    iApp Technology Co., Ltd., Thailand, Mar 2017 - Feb 2018
    • Led the development of the first Thai Treebank (5,000+ entries) with linguists and developers to advance Thai NLP.
    • Developed a syntactic annotation tool for native and web applications, deployed on GCP to support resource development.

Highlight Projects

  • ChindaLLM: LLM-powered Chatbot Platform for Advanced Business Automation (Sep 2024)
    • Led a multidisciplinary team to create a chatbot platform powered by multimodal LLMs with a custom RAG engine.
    • Fine-tuned multimodal LLMs to meet client requirements and developed a graph-based RAG for enhanced retrieval.
  • LLM-based Conversational AI System for Banking Queries (Jul 2024)
    • Contributed to developing an LLM-based agent for general banking queries using TensorRT-LLM and customized RAG.
    • Synthesized data with LLMs to build intent classification guardrail, boosting accuracy by 27% to 92% for banking compliance.
    • Achieved 97.67% QA accuracy, 87.53% recall, and maintained response times under 6.5 seconds.
  • LATTE: Lattice ATTentive Encoding for Character-based Word Segmentation (Jun 2023)
    • Proposed a method using candidate lattices, GNNs, and attention to refine character representations for word tokenization.
    • Integrated Tries with Aho-Corasick to extract candidate characters and words for lattice construction in linear time.
    • Achieved SOTA F1-score (97.7% to 99.4%) across Asian languages: Japanese, Chinese, and Thai.
  • Character-based Thai Word Segmentation with Multiple Attentions (RANLP 2021/Journal of NLP) (Sep 2021/Jun 2023)
    • Proposed a PTM-based word segmentation model with attention across linguistic units (characters, character clusters, subwords, and words), achieving SOTA performance on well-known Thai datasets.
    • Developed a subword tokenizer using SentencePiece and a character-cluster tokenizer optimized for Thai linguistic characteristics.

Selected Projects

  • SpeechFlow: AI-powered Application for Thai-English Transcription, Summarization, and Translation (Dec 2024)
    • Contributed to integrating AI services into an application for seamless Thai-English transcription, summarization, and translation.
    • Led the deployment of the ASR Pro engine on server infrastructure, scaling to support thousands of users.
  • LLM-based Chatbot for Elderly Comfort and Consultation (Oct 2024)
    • Fine-tuned an open-sourced LLM using SFT, DPO, KTO to build a RAG-based chatbot for elderly conversations and support.
    • Designed LLM agents for various tasks, including data synthesis and automatic evaluation.
  • Dual-Stage Face Anti-Spoofing for Active and Passive Liveness Detection (Oct 2024)
    • Led the development of a FAS model with active liveness detection and passive spoofing prevention stages.
    • Achieved Level 1 Presentation Attack Detection certification from iBeta with 0% APCER and BPCER below 3%.
  • ASR Pro: Advanced Context-aware ASR for Thai (Aug 2024)
    • Developed an approach to enhance ASR contextual awareness by integrating LLMs into a fine-tuned ASR model.
    • Reduced WER by 3.12% and improved inference speed by 1.3x than top commercial competitors.
  • Fine-tuning Thai-English TTS Models with Phoneme-level Representations (Aug 2024)
    • Fine-tuned Thai-English TTS models using phoneme-level tokenization, achieving more natural speech than previous models.
    • Contributed Thai-English support to a public TTS repository, extending its functionality with fine-tuned models.
  • Extreme Fine-tuning: A Novel and Fast Fine-tuning Approach for Text Classification (EACL 2024) (Mar 2024)
    • Proposed a fine-tuning approach combining backpropagation with Extreme Learning Machine (ELM) for efficient text classification.
    • Reduced fine-tuning time by up to 74.8% with SOTA-level performance on MELD, IEMOCAP, IMDb, and AG News.
  • LLaVAC: Fine-tuning LLaVA as a Multimodal Sentiment Classifier (Jan 2024)
    • Proposed a method to fine-tune LLaVA for classifying multimodal sentiment labels, incorporating unimodal and multimodal inputs.
    • Outperformed SOTA baselines by up to 7.31% in accuracy and 8.76% in weighted-F1 in the MVSA-Single dataset.
  • A Unification-based Knowledge Graph Construction for Thai Profile Generation from Online Resources (Sep 2023)
    • Constructed a knowledge graph of Thai researchers, using 6+ million entries crawled from online research databases.
    • Designed a semi-supervised method with multi-task learning to extract entities/relations, improving F1-score by 8% over baseline.
  • Simple2In1:A Simple Method for Fusing Two Sequences from Different Captioning Systemsinto One Sequence (Sep 2023)
    • Developed a T5-based generative model for Thai caption fusion, paraphrasing and merging sequences into one.
    • Outperformed baselines by 5.2%, achieving 79% sBLEU and 90% ROUGE-L on a small captioning dataset of 3,168 samples.
  • Multimodal Sentiment Analysis Using Multiple Labels from Different Modalities (Mar 2023)
    • Developed a sentiment analysis model with CLIP, BERT, and RoBERTa, leveraging text, image, and multimodal labels.
    • Achieved up to 2% higher F1-scores than previous models, with 74.1% on MVSA-single and 62.0% on MVSA-multiple datasets.
  • Detecting Fraud Job Recruitment Using Features Reflecting from Real-world Knowledge of Fraud (Mar 2022)
    • Developed a method to classify fake job recruitments using a set of novel features designed to reflect fraudster behaviors.
    • Yielded accuracy of 97.64% for Employment Scam Aegean Dataset (EMSCAD).
  • Public Budget Usage Monitoring System (Feb 2019)
    • Cooperated with an interdisciplinary team to develop a monitoring system that utilizes Scrapy to crawl large-scale unstructured data from government sites, such as procurement and budget portals, for corruption detection in text data. Deployed by two organizations.
    • Developed a text classification method in TensorFlow, with rule-based enhancements, for corruption detection, validated by experts.
    • Obtained Bronze Medal in The 47th International Exhibition of Inventions Geneva

Education

  • Doctor of Engineering, Tokyo Institute of Technology, Japan, 2019 - 2023
  • Master of Engineering, Sirindhorn International Institute of Technology, Thammasat University, Thailand, 2015 - 2018
    • Major: Information and Communication Technology for Embedded Systems
    • Advisor: Prof. Thanaruk Threeramunkong
    • Thesis: A Framework of Thai Treebank Construction and Grammar Refinement
    • Scholarship: TAIST - Tokyo Tech Scholarship
  • Bachelor of Science, Thammasat University, Thailand, 2011 - 2015
    • Major: Computer Science
    • Advisor: Dr. Ratchata Peachavanish
    • Thesis: Web Application for Learning Thai Language using Song Lyrics
    • Activity: President of Computer Science, Faculty of Science, Thammasat University

Skills

  • Technical Skills
    • Programming Languages: Python, C/C++, Rust, Shell Script
    • ML Toolkits: PyTorch/Lightning, TensorFlow, Hugging Face, PyG, OpenCV, Scikit-learn, Spacy, NLTK, TensorRT-LLM, llama.cpp
    • Tools & Technology: Linux, Hadoop/Spark, SQL, NoSQL, Docker, Kubernetes, Elasticsearch, GCP, AWS, Git
  • Languages
    • Thai: Native
    • English: Professional working proficiency
    • Japanese: Limited working proficiency