Ugh, 10,000 rows is honestly pushing it for the base chat interface. It always starts hallucinating once the context window gets crowded. Since you have sensitive data, you really need to be careful about where you upload that stuff. Honestly, I would suggest just biting the bullet and setting up a local environment if you can manage it, but if you need charts by Friday, that might be too much of a learning curve. I have used Julius AI Data Analyst Tool for a few projects and it is decent for quick visualizations, though you gotta double check the math. For the privacy side, make sure you look into their enterprise or individual data handling policies before you dump the whole file in there. Another option is Polymer Search Data Analysis Platform, which is pretty slick for turning spreadsheets into dashboards without needing to code. Just be careful with the free tiers because they might train on your inputs. Whatever you pick, try splitting your data into smaller chunks first. It makes the LLM way more reliable. If the budget is tight, maybe spend some of that fifty on a clean local setup or a secure workspace if your university has a server you can use. Don't stress too much, you will get those charts done in time.