MSc Artificial Intelligence & Machine Learning at the University of Adelaide (2025 to present) and former Analyst at Capgemini. I build practical LLM, RAG and deep-learning systems and care about honest evaluation. Based in Adelaide, Australia, and looking for AI/ML Engineer, Agentic AI, Data and AI Automation internship and graduate roles.
| Project | What it shows | Stack |
|---|---|---|
| rag-cybersecurity-guidelines | Retrieval-augmented generation over cybersecurity guidance with a measured evaluation suite and CI | ChromaDB, Sentence-Transformers, Mistral-7B (Ollama), pytest, GitHub Actions |
| llm-lora-finetuning | Parameter-efficient fine-tuning for classification and dialogue generation | PEFT / LoRA, DistilBERT, RoBERTa, GPT-2, DialoGPT |
| llm-ecommerce-agents | Multimodal and table-reasoning LLM pipeline for e-commerce | BLIP-2, Qwen3-4B, TableLLM-7B, Gradio, 4-bit quantisation |
| blood-cell-classification-deep-res-se-cnn | From-scratch residual SE-CNN, 97.7% validation accuracy, baselines and paper | PyTorch |
| cnn-architectures-from-scratch | MLP vs VGG16 vs ResNet18 comparison | PyTorch |
| llm-trustworthiness-bias-evaluation | Bias, toxicity, privacy and jailbreak evaluation with ANOVA | Mistral-7B, Detoxify, SciPy |
| rlhf-dpo-reward-model-gpt2 | SFT, reward model and DPO alignment | TRL, Transformers |
| vae-mnist-pytorch | Variational autoencoder from first principles | PyTorch |
| pca-kmeans-lightgbm-classification | PCA, K-Means from scratch, 92.9% 7-class LightGBM | scikit-learn, LightGBM |
| woolworths-trust-recovery-dashboard | Executive dashboard storytelling | Tableau |
More repos: PyTorch pipelines, diabetes MLP, Keras CNN and Transformer, regression from scratch, cybersecurity risk assessment, multimodal disease prediction. Each repo's README states its results and its limitations.
Languages: Python, Java, SQL (plus JDBC) ML / DL: PyTorch, TensorFlow / Keras, scikit-learn, LightGBM, Hugging Face Transformers, PEFT / LoRA, TRL GenAI: RAG, prompt engineering, LLM evaluation, fine-tuning, quantisation, vector search (ChromaDB) Data and tools: pandas, NumPy, Tableau, Git / GitHub Actions, pytest, Jupyter / Colab
Repos are university coursework and personal projects. Course datasets are not redistributed.