This week, I delved deeper into the two projects I am a part of. I first went through all the pre-existing documents and discourse for the two projects, so that I could establish a strong foundation that I can refer back to as I move forward with my work. After this, I then focused on the LLM project for the rest of the week. Currently, this project is in the proof of concept (PoC) stage, meaning that the data team is experimenting with the extent of the program’s capabilities, and which of the brainstormed ideas are actually feasible. I was also able to meet with the project’s data scientists, Yasmin and Enrique, to discuss their experiences during this exploratory stage, as well as to gain clarity on which coding languages and models the LLM will utilize. Combining this information with my findings from last week’s literature review on bias in AI, I began to brainstorm potential outlets of bias in the LLM, where I focused on areas such as language and dialect nuances, code definitions, and choice of data sources. At the end of the week, I presented my findings to my project manager, Zinnya, where we discussed how to address testing my hypotheses.

The start of the week posed challenges due to my project manager's absence at a conference, which meant our regular check-in meetings couldn't take place. Without having the ability to message her, I started the week unsure of how to start and move forward with my work. However, I reached out to my team member, Adjani, who was able to help me solidify my tasks, all while being super kind and approachable. Presenting my findings also went really well, as I received good feedback from Zinnya on my thoughts and ideas. 

Throughout this week, I've grown more comfortable with DPA and developed stronger relationships with my team members. While I continue to encounter uncertainties, particularly during coding language discourse, I am picking things up quickly. Next week, I plan to continue researching my hypotheses to determine their legitimacy as outlets for bias. This involves interviewing colleagues, researching related discussions, and testing other LLMs like ChatGPT.