Lead Consultant at the Graduate Quantitative Methods Center, UC Riverside — workshops and one-on-one methods consulting for graduate students across the university.
GradQuant is UC Riverside’s campus-wide center for quantitative and computational methods training. As Lead Consultant I work with graduate students from every school on campus — on their own data, at whatever stage they have reached.
Getting set up in Colab, then the essentials: variables and data types, strings and f-strings, lists, tuples and dictionaries, indexing and slicing, functions and modules, and plotting with matplotlib. The session closes with a mini project that pulls CPI and unemployment from FRED, computes year-over-year inflation, and fits a Phillips curve.
Why time series breaks the OLS assumptions, and what to do about it. Trend, seasonality, cycles, shocks and structural breaks; stationarity, unit roots and the ADF test; seasonal decomposition; reading ACF and PACF to choose lag orders; estimating ARIMA and SARIMA, validating with Ljung–Box, and forecasting out of sample. Worked on U.S. real GDP and the airline passenger series.
Four strategies for forecasting with many correlated predictors, compared on a simulated macro dataset with three latent factors and sixty indicators: PCA and factor models (compression), partial least squares (supervised compression), Lasso (sparse selection), and neural networks (representation learning). Covers the estimation problem behind each method and how to evaluate them with out-of-sample RMSE, MAE and R² against an AR(1) benchmark.
First the mechanism — tokens, embeddings, positional encoding, attention, multi-head attention, encoder and decoder blocks, and what pretraining and post-training do to a model’s behavior. Then research practice: when an LLM beats a dictionary or a fine-tuned classifier and when it does not, calling an API reproducibly, grounding prompts in source text, and validating model output against hand-coded samples — demonstrated by scoring FOMC statements for hawkishness.
Thirty-six scheduled appointments and twelve drop-in sessions. The scheduled ones drew students from CNAS, CHASS, BCOE, the School of Public Policy, the School of Medicine and the Graduate School of Education — entomology, plant pathology, psychology, political science, computer science, materials science, public policy, economics and more. Twenty-one were doctoral students, nine master’s students, five postdoctoral researchers; fifteen were international students.