# <<cv.name>>'s CV
((* if cv.label *))
*<<cv.label>>*
((* endif *))

((* if cv.phone *))
- Phone: <<cv.phone|replace("tel:", "")|replace("-"," ")>>
((* endif *))
((* if cv.email *))
- Email: [<<cv.email>>](mailto:<<cv.email>>)
((* endif *))
((* if cv.location *))
- Location: <<cv.location>>
((* endif *))
((* if cv.website *))
- Website: [<<cv.website|replace("https://","")|replace("/","")>>](<<cv.website>>)
((* endif *))
((* if cv.social_networks *))
    ((* for network in cv.social_networks *))
- <<network.network>>: [<<network.username>>](<<network.url>>)
    ((* endfor *))
((* endif *))

# <<section_title>>
## **University of New South Wales***Sydney, Australia*

*Feb 2027 – Dec 2029*

*Master of Statistics*



## **University of New South Wales***Sydney, Australia*

*Feb 2021 – Sept 2025*

*Bachelor of Computer Science (AI), Minor Mathematics*



# <<section_title>>
**Deep Learning:** PyTorch, TensorFlow, HuggingFace Transformers, nnU-Net, U-Net, CNNs, LLMs, RAG

**ML Engineering:** Python, CUDA, HPC (H200/A200), Distributed Training, Model Optimisation, Hyperparameter Tuning

**Computer Vision:** Semantic Segmentation, 3D Medical Imaging, Data Augmentation, Preprocessing Pipelines

**LLM & GenAI:** LangChain, LangGraph, LlamaIndex, smolagents, OpenAI API, Prompt Engineering, Fine-tuning, RAG

**MLOps & Cloud:** Docker, AWS, Azure, Git, Experiment Tracking, Model Deployment

# <<section_title>>
## **[Biomedical Semantic Segmentation](https://abaj.ai/projects/dl/kits19)**

*KiTS19*

Production-grade kidney and tumour segmentation from 3D CT scans on official KiTS19 leaderboard.

- Achieved 0.9129 Dice score (ranked #57 globally) by implementing nnU-Net with systematic hyperparameter optimisation on H200 GPUs.

- Reproduced state-of-the-art 3D medical imaging pipeline end-to-end; published technical report with ablation studies comparing 2D vs 3D architectures.



## **Full-Stack LLM Chat Application**

Production-ready conversational AI with RAG pipeline, streaming responses, and persistent context.

- Architected full-stack chat application using TypeScript, Bun, Svelte, and Prisma ORM with PostgreSQL; integrated OpenAI GPT API with streaming.

- Implemented security middleware with rate limiting, response caching, and structured logging; validated with Zod schemas and comprehensive test suite.



## **[Neural Networks: Zero to Hero](https://github.com/abaj8494/10khrs-ai-ml-dl)**

*Karpathy*

From-scratch implementations of neural networks culminating in a GPT language model.

- Built micrograd (autograd engine), makemore (character-level LM), and nanoGPT following Andrej Karpathy's curriculum.

- Implemented backpropagation, attention mechanisms, and transformer architecture from first principles in pure Python/PyTorch.



## **[Dead Tree Segmentation](https://abaj.ai/doc/publications/dead-tree-seg.pdf)**

*Kaggle*

Semantic segmentation of live/dead trees from satellite imagery with severe class imbalance.

- Achieved 0.71 IoU on 97% imbalanced dataset by implementing U-Net from scratch (31M parameters) with Dice + Focal loss combination.

- Parallelised training pipeline on CUDA-enabled HPC; applied morphological post-processing for boundary refinement.



## **[Augmenting Infrastructure](https://abaj.ai)**

*Monorepo*

Technical knowledge platform with ML/AI research notes, graduate-level mathematics, tutorials, and paper summaries.

- Architected full-stack content platform with Hugo static generation, Emacs org-mode authoring, and bidirectional linking across 1,073+ pages.

- Built literate programming environment with org-babel for executable Python/Julia notebooks, replicating Jupyter workflows with version control.



# <<section_title>>
## **AWS Solutions Architect (SAA-C03)**

*Jan 2026 – Feb 2026*

*Amazon Web Services*



## **AI Agents Course**

*Jan 2026 – Jan 2026*

*Hugging Face*