Reinforcement Learning-Based Al for Sustainable Stock Trading System in the Stock Exchange of Thailand

In 2024, I present an application of reinforcement learning (RL) techniques to the domain of stock trading on the Thai Stock Exchange (SET). Unlike traditional approaches that rely on memorizing historical data, My RL-based method enables an AI agent to actively learn and adapt through iterative interactions with the trading environment. I define the state space, action space, and reward functions to enhance the agent's learning process. The state space captures relevant market indicators and stock-specific features, while the action space allows the agent to make buy, sell, or hold decisions for each trading period. The reward function is designed to maximize long-term portfolio returns while accounting for transaction costs and risk management considerations. Through continuous exploration and exploitation, the RL agent learns an optimal trading strategy tailored to the dynamics of the SET market. My experimental results from backtesting demonstrate the efficacy of the proposed approach, outperforming Thai SET index.

AI solution for predicting flood

In 2023, Flooding is a severe natural hazard that can cause substantial loss of life and property damage, To address this critical issue, I present a approach for flood prediction by leveraging water level data collected from stations across the country.
I focus on predicting water levels at various stations across Thailand to forecast potential flooding events. By analyzing the data collected from these stations, I can predict water levels for the next three days. If the predicted water level exceeds the bank threshold, it indicates a high likelihood of flooding at that station. My prediction model achieves an accuracy of over 98%, providing a reliable tool for flood risk management and mitigation. You can view it at the following link

AI-Driven Sales Forecasting for Retail Stores

In 2023, As the team leader and primary AI programmer, I led a project to forecast retail store sales using AI models, including Prophet by Meta and AutoRegressive Integrated Moving Average (ARIMA). Over a 2-month period, we aimed to improve sales prediction accuracy by integrating holiday effects and seasonal trends, analyzing historical sales data to identify key patterns. My responsibilities included developing and implementing these models, ensuring their accuracy and reliability. The project resulted in significantly more accurate sales predictions, enabling the retail stores to optimize their inventory and marketing strategies effectively.

AI-controlled Mouse and Keyboard for People with Arm Disabilities

In 2022, In today's digital age, computer accessibility remains a significant challenge for individuals with arm disabilities, hindering their ability to interact effectively with technology. I present a Python-based application that aims to bridge this gap by enabling mouse and keyboard control through the intuitive interpretation of head poses and eye blinks.
Leveraging the power of Mediapipe, a cutting-edge Google library specialized in facial landmark detection, My application employs computer vision techniques to track and interpret the user's head movements and eye blinks. By mapping these natural gestures to mouse and keyboard commands, I provide an innovative and intuitive solution for individuals with limited arm mobility to navigate and interact with computers seamlessly. [github]

Stock Investment AI based on Long short-term memory (LSTM)

In 2022, At the core of my AI model lies a robust LSTM network architecture,By ingesting a comprehensive dataset of historical stock prices [high, low, close], trading volumes, and relevant market indicators, my model learns to identify intricate patterns and relationships, enabling it to make accurate predictions of future stock trends within the SET market. My experimental results from backtesting demonstrate the efficacy of the proposed approach, outperforming Thai SET index.

NextGen AI Camp

In 2024, Out of nearly 3,000 participants, an initial assessment was conducted over two days, covering mathematics, coding skills, and fundamental AI. From this, 300 individuals were selected. Over the next two months, we were assigned weekly AI challenges and tests, with scores accumulated from weekly submissions. After the two months, the top 30 highest-scoring individuals were chosen to participate in an on-site AI Hackathon, which lasted four days. From there, the group was narrowed down to 10 finalists, and I ranked 2nd among them. This achievement granted me admission to the Computer Engineering program at King Mongkut's Institute of Technology Ladkrabang (KMITL).

TA for an Introduction to Machine Learning course at KVIS

I've developed and delivered a curriculum that not only educates but also inspires students to apply AI in real-world scenarios.

1. Introduction to OpenCV and Mediapipe
2. Real World Applications(OpenCV) [github]
   • Counting Objects
   • Scanning Passport and Receipts
   • Shape Detection
   • Bubble Sheet Score
   • Ball Tracking
   • Measuring Size
   • Face Detection
3. Real World Applications(Mediapipe) [github]
   • Face Detection
   • Image Classification
   • Object Detection
   • Eye Blink Count
   • Playing Games
4. Image Classification [github]
   • CNN (Convolutional Neural Networks)
   • Training with custom CNN
   • Transfer learning with ResNet34,YOLOv8
5. Object Detection [github]
   • COCO dataset
   • YOLO (You Only Look Once) algorithm
   • Real-time Object Detection using pretrained models like RT-DETR and Transfer learning
   • Application for car license plate detection and speed estimation using YOLOv8 and OpenCV

President, Artificial Intelligence (AI) Club, KVIS

In 2023, In my classroom, I've developed and delivered a curriculum that not only educates but also inspires students. The curriculum covers a range of topics designed to ignite curiosity and practical skills in the field of artificial intelligence, Time-Series Analysis, Computer vision, Transformers and Real-World Applications.

1. Introduction to Python
2. Introduction to Machine Learning
3. Machine Learning for Data Science
4. Machine Learning for Image Classification
5. Machine Learning for Object Detection
6. What Is a Transformer Model?

Website serves as a centralized platform for adding, storing, and searching for Thai researchers

The website functions as a centralized platform designed for adding, storing, and discovering Thai researchers. It aims to simplify access to information about researchers' work throughout Thailand, offering a comprehensive database that enables users to explore and engage with the academic community.

International Exchange Program, Korea Science Academy of KAIST (KSA)

KSA is renowned for its excellence in science and technology education. Studying at KSA grants me access to state-of-the-art research facilities and a rigorous academic curriculum that surpasses traditional educational standards.

AI Researcher in Super AI Engineer Season 2