migraine was really bad again today. is it ia lack of quality sleep? stress? braces? dehydration? screens? so many potential causes. i got home and just laid on the floor and was knocked out for an hour.

first time cooking japanese beef curry today with my mueller 6 qt enamled cast iron dutch oven. the entire process took 3 hours end-to-end. taste wise, the flavor could be better, and my beef wasn't soft enough. i think it needs curry powder and bay leaves. and i shouldn't use sirloin for my beef, but chuck roast.

spoke to someone working at stripe as a DA.

a few takeaways below.

3 components of work

• define key metrics to track
• running AB tests, interpret results, provide product strategy and recommendations
• deep dive analysis, sizing opportunity, i.e. forecasting cost for next year, causal inference

writing culture at stripe

• always write a doc when making a decision, get opinions from others, have a review session
• compared to traditional powerpoint culture, no need to spend time on making things look nice

stripe vs MSFT

• stripe is growing, still smaller than FAANG, fast moving culture, cares more about startups and smaller business
• microsoft is large, things move slower. more solid structure for doing things
• at stripe, you just do things, there's good and bad side
• MSFT has a lot of usage data, how people use office and teams to run analysis to better serve users. ex: how to acquire more seats (more subscriptions for office products)

stripe DS vs DA

• DS runs more ML models than DA
• DS would deploy a model
• MLEs would do even more engineering type of work, setup envs, MLOps
• experiments is fun, you can learn something totally different from what you expected

experimentation at stripe

• coordinate with eng where to put code,
• do forefront analysis for how much samples needed, metrics to track,
• AA testing to ensure experiment setup is right to not waste time to put in bad data
• it's more manual, microsoft is more streamline to run experiments

experimentations are moments of truths

• experiment results gives you learnings for product roadmap
• i.e. results showed cost but also decrease in conversion rate, run further tests to decide action plan. care about cost or customer more?

advice on career

• look at people ahead of you and where they landed
• gather data points
• it's your choice in the end
• you have to try, sometimes you never know, you have to do it and see if you like it or not
• it's a luck thing to find what you love