Self-Introducing Chatbot using RAG with Local LLMs
-
Leveraged LLaMA3 with RAG: Implemented the LLaMA3 model locally, using RAG to integrate PDF data (CV and personal statement), for personalized responses.
-
Developed User Interface: Created an interactive web page allowing users to input questions and receive real-time responses.
Demo
Brain-based Reinforcement Learning for Decision Making
(Code pending)
This is my PhD research, focuses on using a Multi-Armed Bandit (MAB) model to complete three tasks inspired by insect brains. For detailed insights and findings, please refer to my publications.
-
Identify the optimal learning strategy that enhances insect agent survival in hazardous environments with varying memory types.
-
Evaluate the effectiveness of different synaptic plasticity forms during learning.
-
Determine the most efficient decision-making process for foraging using a contextual bandit approach.
Room Cleaning Robot Based on Deep Neural Network
-
Used Deep Neural Network and HSV colour detection into the Computer Vision system and achieved real-time Object detection and classification.
-
Achieved the distributed computing of the robot control system based on Robot Operating System, desiend the robot manipulation system includes the control and motion planning of the robot arm, gripper and mobile base.
Simulation
Robot demonstration
Python-based Numerical Techniques in Solving Black-Scholes equation
-
Resolved the Black-Scholes partial differential equation by applying Monte Carlo model and the Finite Difference Method (FDM).
-
Implemented Explicit, Implicit and Crank–Nicolson FDM algorithms with Python, applied Numpy to finish decompositions.
-
Analyzed the stability and convergency of the models, compared the resulting accuracy and computational cost of the FDM models against the Monte Carlo model.
A Raspberry Pi and TensorFlow powered self-driving robotic car
A robotic car who is able to navigate the lane autonomously by using deep convolutional neural networks (CNN), and detect various traffic signs and pedestrians on the road by Single Shot multibox Detector (SSD).
A cute Pusheen cat chatbot based on Andriod
-
A cute chatbot based on Java and Android Studio, allows both voice and text input, which allows users to chat via text and voice messages and manage chat histories.
-
Applied Simsim API, Google Firebase API and Andriod Speech to relize the voice and text conversation.
