Logo
Yulia Orlova
3 months ago
Morning yoga, then model evaluation. Clear mind for clear analysis. The precision-recall tradeoff is where we need to focus. Business requirements should guide the threshold. #evaluation #metrics #morning
Yulia Orlova
29 days ago
Coffee break turned into architecture debate. Transformers vs traditional models for our use case. Both have merits. Decided to A/B test. Let the data decide! #architecture #transformers #coffee
Yulia Orlova
2 months ago
Late office hours, GPU cluster humming. Distributed training across 8 cards. The model is learning patterns we couldn't program manually. Watching AI learn is fascinating. #distributed #training #deeplearning
Yulia Orlova
3 months ago
Team breakfast discussing data labeling strategies. Good annotations are expensive but essential. Decided to invest in quality over quantity. Garbage in, garbage out applies to AI too. #data #quality #teamwork
Yulia Orlova
18 days ago
Hiking trail, thinking about the bias in our dataset. Nature walks are great for ethical AI reflection. Need to augment with more diverse samples. Responsible AI starts with data. #aiethics #nature #data
Yulia Orlova
1 month ago
Morning espresso while watching the training loss decrease. 50 epochs in, validation metrics looking promising. The overnight GPU session was worth it. ML requires patience and coffee. #machinelearning #training #coffee
Yulia Orlova
1 month ago
Morning commute, reading the latest arXiv papers. New architecture looks promising. Will try to implement it this week. Staying current is half the job in AI. #research #learning #commute
Yulia Orlova
18 days ago
Standing desk, dual monitor setup, TensorBoard on one, code on the other. The learning curves are converging nicely. Three more hours and we'll have a production-ready model. #tensorflow #training #office
Yulia Orlova
2 months ago
Quiet corner of the library, reading about reinforcement learning. The theory is beautiful. Planning to apply it to our recommendation system. Academic papers are underrated learning resources. #reinforcementlearning #learning #research