TA for the computer architecture course

• Created pipelines autonomous deployment.• Created Infrastructure code to create deployment resources.• Optimized resource usage in terms of memory and computation.• Optimized costs for the deployments.

• Fine-tuned several LMs to mask sensitive information in user messages sent to chatbots in English, Dutch, Standard and Cairene Arabic. This includes over 27 entities as defined by the GDPR. Our best model had an F1 of 92.5%.• Fine-tuned LMs for extractive QA in English, Dutch and Arabic. Ported 7B and 13B Llama 2 models to run in-house with GPTQ optimization for efficiency and prompt-engineered it for abstractive QA and validated both using Squadv1 dataset.