Featured Projects

📊

FinGraphRAG Portfolio Risk & Market Analysis

A comprehensive GraphRAG system for portfolio risk and market analysis, combining vector search (Qdrant) with knowledge graph reasoning (Neo4j) to provide explainable financial insights.

GraphRAG Neo4j Qdrant FastAPI React Python Docker
📈

FutureVest Stock Analysis | CSSE6400

A web-based stock analysis platform helping investors value stocks and discover growth opportunities with screening, watchlists, alerts, and technical analysis tools.

React TypeScript Flask Python PostgreSQL AWS Docker
🤝

Co-Founder Matching UQ Ventures | DECO3801

A mobile app connecting entrepreneurs with potential co-founders using a multi-factor matching algorithm evaluating skills, interests, goals, and collaboration style.

React Native TypeScript Flask Python Supabase Docker
🔍

DF-P2E: Deepfake Detection CSIRO Research

Created a multimodal deepfake analysis system integrating CNN/Vision Transformer detection with explainable AI techniques (SHAP, LIME) and LLM-powered forensic insights using Llama 3.1.

PyTorch Vision Transformers CNNs SHAP LIME CLIP Llama 3.1 Hugging Face
🤖

Multi-Agent Customer Support Triage CrewAI

Built a multi-agent system for customer support triage using CrewAI framework with LangSmith for observability. Features ticket classification, policy retrieval, response drafting, and compliance enforcement.

CrewAI LangSmith Multi-Agent Python LLMs
📚

arXiv Self-Correcting RAG Pipeline LangGraph

Built a multi-agent arXiv self-correcting and adaptive RAG pipeline that intelligently retrieves, processes, and synthesizes research papers with self-correction mechanisms.

RAG Multi-Agent LangGraph Python LLMs
📊

VideoSpec & Creator-Brand Matching LINQIA

Built multimodal video-analysis platform with Gemini Vision API and developed algorithms analyzing millions of creator profiles with NLP and computer vision for brand-creator matching.

Python Gemini Vision API AWS Nova NLP Computer Vision LLMs Streamlit
🤖

AIM AI AI Interviewer

Developed an AI-powered interviewer chatbot by fine-tuning Mistral-7B using QLoRA, achieving 18% improvement in Rouge-1 scores with integrated speech-to-text processing 4x faster than original Whisper.

Mistral-7B QLoRA Fine-tuning Faster Whisper CTranslate2 Python

Energy Transition Analysis UQ Research

Developed ML tools for energy-project trend analysis using AEMO data, achieving ~15% reduction in estimation errors for energy transition model calibration.

Python Machine Learning Deep Learning Data Analysis LiDAR
💰

QuickSettle Westpac Hackathon

Co-developed a group payments app with shareable links and automated reminders, presented at the Gold Coast Engineering Hub as a Hackathon Finalist.

FinTech App Development Hackathon
🧠

Alzheimer's Disease Classification Vision Transformer

An MRI-based Alzheimer's disease classification project exploring the use of Vision Transformers (ViT) for medical image analysis. The system implements patch extraction, linear embeddings, positional encoding, and multi-head self-attention from scratch using TensorFlow/Keras, and is trained and evaluated on the ADNI dataset.

TensorFlow Keras Vision Transformers Python Medical Imaging

Want to learn more about my work? Check out my GitHub profile or connect with me on LinkedIn!