About
I am Saksorn (Harry) Ruangtanusak, an applied machine learning researcher at SCB DataX in Bangkok, working with the Typhoon team on language models and agents.
My work spans LLM post-training, reasoning-model evaluation, tool-using agents, multilingual agent benchmarking, retrieval, and production AI systems. Previously, I worked at SCBX R&D, CJ MORE, and AI and Robotics Ventures.
I studied Mechanical Engineering (2nd Hons.) at KMUTT, where I worked on neutronics under the supervision of professor Yanin Sukjai.
Through the SCBX–Stanford HAI Corporate Affiliate Program, I collaborated with professor Michael S. Bernstein and Jure Leskovec to develop AI research and use cases for the SCBX Group.
Currently, I am interested in how LLM agents can extend human capabilities while remaining reliable, controllable, and grounded in real-world workflows. I am especially interested in long-horizon human–agent collaboration and oversight.
News
- Aug 2026SEATauBench was accepted to Findings of EMNLP 2026.
- Jul 2026SCBX Launch CORA, SCBX's AI analyst for complex Thai financial documents, originated from my research on RAG (DoTA-RAG), published at the SIGIR LiveRAG Workshop.
- Jul 2026Talked at Chiang Mai University, Faculty of Economics: Reimagine Your Workflows using Typhoon with Coding Agent.
- Jun 2026SEATauBench, our multilingual benchmark for tool-using agents in low-resource Southeast Asian languages, released on arXiv.
- Apr 2026Featured by SCBX: Using Agent Simulation to Improve Marketing Decision Making.
- Nov 2025Techsauce covered our 5th-place finish in the ACM SIGIR 2025 LiveRAG Challenge, the highest-ranked industry team among 40 teams.
- Oct 2025Co-authored SCBX AI Outlook 2025 Vol. 2, covering the shift from scaling models toward capable and responsible AI.
- Jul 2025The Typhoon team's A* AI research collaborations were featured by Techsauce.
- May 2025Talked at Super AI Engineer Season 5 with Peerawat (Boat) Rojratchadakorn: From SuperAI Engineer to SCBX: Journey into AI Fintech Innovation.
- 2025DoTA-RAG presented orally at the SIGIR 2025 LiveRAG Challenge.