AI Research Intern at Truuth Pty Ltd

Independently designed and delivered bespoke AI solutions for document fraud detection — scoped requirements, selecting models, building pipelines, and deploying on AWS SageMaker with minimal supervision. Built multi-step agentic LLM workflows (GPT-4, LLaMA, AWS Bedrock) to automate internal review, metadata extraction, and analytical reporting. Integrated generative AI tools into existing business processes: automated document summarisation, structured data extraction from unstructured inputs, and anomaly flagging for compliance workflows. Developed and evaluated ML classification models using statistical image forensics (ELA, ViT, CNNs) to detect forged document regions; iterated on model performance through rigorous evaluation pipelines. Collaborated cross-functionally with engineering and product stakeholders in an agile environment — gathering requirements, iterating on solutions, and documenting technical processes for handover.