Saksorn Ruangtanusak

Saksorn Ruangtanusak

AI Researcher | Also known as "Harry"

Bangkok, Thailand

(+66) 83-062-4043

saksorn.ruang@gmail.com

Research Interests

Large Language Models & Agentic Systems

Multilingual agents, tool use, planning and evaluation, simulation.

Retrieval-Augmented Generation (RAG)

Dynamic retrieval, robust evaluation, hallucination reduction.

Speech & Multimodal AI

Robust automatic speech recognition and vision-language models.

Experience

AI Researcher

Oct. 2024 - Present

SCBX - R&D and Innovation Lab | Bangkok, Thailand

  • Personalized Ads with Agent Simulation (SCB x Stanford HAI): Leading LLM-based ad preference simulation and persona agents for SCB investment products, in collaboration with Stanford HAI (advised by Prof. Michael Bernstein). Designed large-scale SCB ad-variation datasets and evaluation pipelines.
  • Agentic AI for Investment (InnovestX x BCG): Co-developing agentic workflows that connect investment research and customer journeys into explainable recommendation chains for InnovestX employees, in collaboration with BCG.
  • Robust ASR Across SCBX Subsidiaries: Improving internal ASR robustness speech from multiple SCBX entities. Curated training and evaluation sets.
  • Core contributor to SCBX AI Outlook reports (Vol. 1 & 2), positioning SCBX R&D as an AI research lighthouse inside the group and summarizing global frontier trends.
  • Research Mentorship: Advised an SCBX intern whose work on Thai semantic end-of-turn detection was accepted to ICSEC 2025, the biggest computer science conference in Thailand.

AI Researcher

Mar. 2024 - Oct. 2024

CJ MORE - Technology Innovation Lab & Decision Intelligence (TILDI) | Bangkok, Thailand

  • Developed LLM and RAG systems for retail chatbots and knowledge management, including customized document parsers and hierarchical HyDE-style retrieval.
  • Built internal tools to help non-technical teams access data and analytics, reducing reliance on data engineers.
  • Experimented with efficient fine-tuning of vision-language models (VLMs) and small language models (SLMs) for on-premise deployment.

Deep Learning Data Scientist

Jun. 2022 - Mar. 2024

AI and Robotics Ventures (ARV Bedrock) | Bangkok, Thailand

  • Developed 2D image (classification, detection, segmentation, OCR) and 3D point cloud models for smart tax and smart city solutions using MMS survey cars.
  • Designed a 3D pipeline from scratch for size and location extraction on point clouds; deployed containerized APIs and inference pipelines on AWS.
  • Worked on few-shot segmentation core services for projects with very limited labeled data (1-100 shots).
  • Built analytics for Thai public health and smart city platforms using ITIC datasets, combining geospatial deep learning and NLP clustering of POIs.

Research Assistant - Neutronic Analysis

Sep. 2020 - May 2022

King Mongkut's University of Technology Thonburi (KMUTT) | Bangkok, Thailand

  • Performed Monte Carlo neutron transport simulations and neutronic analysis for next-generation boiling water reactors.
  • Co-authored international publications on transuranic waste incineration and reactor design with Prof. Yanin Sukjai.

Super AI Engineer Season 2 (Gold Medal) - Apprentice

Jan. 2022 - Apr. 2022

Artificial Intelligence Association of Thailand | Bangkok, Thailand

  • Solved 8 diverse AI hackathons (NLP, CV, signal, robotics, IoT, data science); several top-3 finishes in tasks such as Thai word tokenization, EEG classification, and robotics.
  • Led the motion-control team for a first-aid robot, achieving top performance using PID control with MATLAB.

Selected Publications

SIGIR 2025

DoTA-RAG: Dynamic of Thought Aggregation RAG

S. Ruangtanusak et al.

Proc. of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval (LiveRAG Challenge), 2025. Oral

EMNLP 2025

Talk Less, Call Right: Enhancing Role-Play LLM Agents with Automatic Prompt Optimization and Role Prompting

S. Ruangtanusak, P. Taveekitworachai, K. Pipatanakul

EMNLP 2025 WordPlay Workshop, 2025. Spotlight

PHYSOR 2022

Verification of OpenMC Neutronic Code for Transuranic Waste Incineration in Reduced Moderation Boiling Water Reactor

S. Sansawas, S. Ruangtanusak, Y. Sukjai

International Conference on Physics of Reactors (PHYSOR 2022), American Nuclear Society, 2022.

Awards & Hackathons

1st Prize - Typhoon AI Hackathon 2024

SCB10X | Jun. 2024

Built "MANage In The Middle (MITM)", a personalized meeting summarizer and feedback tool using ASR and Typhoon 1.5x 70B as LLM agents.

1st Prize Financial Winner - Bangkok AI Hack 2023

SCB10X | Nov. 2023

Built "PlayVest", a generative AI financial adviser platform integrating a simulated trading system, guardrails, intent classifier, RAG, and text-generation models within 9 days.

Global Open-Source AI Game Jam

Hugging Face | Jul. 2023

Built a visual novel game, "Expand the Love", with open-source AI within 2 days. Utilized Stable Diffusion, SRGAN, GPT-4, and Unity.

1st Prize - Autonomous Anti-Drone System

AI and Robotics Venture | Jun. 2023

Created "The Dispatcher", an autonomous anti-drone system. Applied ML techniques for drone detection, classification, and tracking at 60 fps on Jetson Xavier.

Skills & Education

Technical Skills

Programming

Python (Advanced) SQL (Advanced) Bash (Advanced) JavaScript (Basic)

ML / DL

PyTorch Hugging Face OpenCV GeoPandas MLflow Scikit-learn TensorFlow Keras

MLOps & Cloud

Docker Conda AWS GCP (Basic) Dagster Looker Studio

Domains

NLP LLMs RAG Agentic Systems Computer Vision (2D/3D) Geospatial Deep Learning ASR

Soft Skills

Project Management Requirements Gathering Non-technical Communication Mentoring

Languages

Thai (Native) English (Intermediate, IELTS 6.5)

Education

B.S. in Mechanical Engineering

2nd Class Honors

King Mongkut's University of Technology Thonburi (KMUTT)

2018 - 2021 | Bangkok, Thailand

  • Research Assistant in the Nuclear, Thermal, and Energy Engineering Lab under Prof. Yanin Sukjai.
  • Received a university scholarship; completed extra coursework in ML, Big Data, Image Processing, and Robotics.