This entry is different: not a retrospective but a plan. I am in the eighth of nine terms of Plan B, two away from graduating in May 2027.
Architecture for Large Data Volumes
Distributed storage, MapReduce and Spark, columnar formats, partitioning, and the design of systems that do not fit on one machine. I arrive with an advantage: at work I live inside BigQuery, which is exactly this with a different interface. What I want to get is what work does not teach you: why the system is designed this way, what it sacrifices, and what I would do differently if I designed it.
Data Mining and Analysis
Association rules, clustering, dimensionality reduction, text analysis. It is the “exploratory” course of the degree: finding structure where nobody said there was any. My particular interest is text, because of last term’s Bag-of-Words and because of WhisperFlow, my offline dictation tool: I want to understand better the language pipelines I already use.
Multivariate Methods
The multivariate normal, principal components, factor analysis, canonical correlation, discriminant analysis, multidimensional scaling. Principal components is the bridge between second-term Linear Algebra (eigenvalues) and sixth-term machine learning: reducing dimensions is diagonalizing the covariance. I like it when a course closes a circle another one opened three years earlier.
Applied Data Science I
An integrative project with a real partner, over two terms. Since this fall the case developed here can replace the thesis as the path to graduation. It is the course in which everything before has to show up at once: question, data, model, causality, ethics, communication. I am still choosing the problem; I want it to have public data and a question that matters to someone.
Public Policy II
The second part of the applied public-policy topic, now with last term’s causal-inference tools. Evaluating an intervention with difference in differences, and explaining it to someone who does not know what a counterfactual is.
What comes next
The ninth term brings Bayesian Statistics (the course I look forward to most), Applied Data Science II, Professional Communication and three electives; the candidates include Time Series, Stochastic Processes and Statistical Learning. And the 480 hours of social service remain.
What I want to take with me
That the last stretch of the degree is not a list of passed courses but a project I can still talk about five years from now. This log is, in part, a way of holding myself to that.