projects

Research and Developer Project Experience

Capstone Projects

SATU: Agentic Physical AI for Human Friends (AY2025)

Team: Kirawut Chalermkitpaisan, Krittin Sakharin, Natcha Rungruang, Natwasa Manomaiwiboon
Advisor: Asst. Prof. Dr. Sarucha Yangyong

An intelligent, autonomous mobile robot designed to serve as a receptionist and human assistant. Built as a research agentic system, it integrates custom tools and Model Context Protocol (MCP) to provide the LLM with a physical, human-like embodiment capable of natural Thai conversation, real-time human detection, and safe autonomous navigation.

  • 🧠 Brain: Typhoon2-70B & Typhoon2-8B
  • 🗺️ Navigation: ROS 2 Jazzy, Nav2 Stack, RPLiDAR
  • 💾 Database: Milvus Vector DB
  • 🗣️ Speech: SileroVAD, Typhoon 2 Audio
  • 👀 Vision: Jetson Nano, SCRFD, ArcFace
⚡ Response: ~2.55s (p50) 🛡️ Safety: 100% Success Rate ▶️ TikTok Demo
SATU: Agentic Physical AI for Human Friends

RISA: A Human-Engaging Agentic Robotic Interactive Assistant (AY2024)

Team: Mrs. Chatchaya Miyamoto, Mr. Phoomiphat Kittisuphat, Mr. Siripong Boonsiri, Mr. Sivakorn Samorkam
Advisor: Dr. Sarucha Yanyong

An interactive robotic assistant tailored for guided lab tours, built from scratch using a prototype robot from the RoboCup Japan Open 2024, Logistics League. RISA is a research agentic framework that leverages custom tools and Model Context Protocol (MCP) to give the LLM a physical presence. It combines Retrieval-Augmented Generation (RAG) with facial recognition for personalized, highly accurate human engagement.

  • 🧠 Brain: Llama 3.2 & Mistral
  • 🗺️ Navigation: ROS1 Noetic, LiDAR, IMU
  • 💾 Database: ChromaDB + MongoDB
  • 🗣️ Speech: OpenAI Whisper, Botnoi Voice API
  • 👀 Vision: YOLOv5s, DeepFace
⚡ Response: Hybrid RAG + LLM 🛡️ Safety: 80% Avoidance Success
RISA: A Human-Engaging Agentic Robotic Interactive Assistant

Research and Developer Projects