I work as an engineer at Kindred AI with a focus on applied machine learning. I am particularly interested in ML applications related to interactive information retrieval, Bayesian optimization, and robotics. More generally, I'm interested in the principled design, implementation, evaluation and optimization of machine learning systems.
resume | twitter | github | linkedin | quora
Many real-world engineering problems rely on human preferences to guide their design and optimization. PrefOpt is an open source package for simplifying sequential optimization tasks that incorporate human preference feedback. Our approach extends an existing latent variable model for binary preferences to allow for observations of equivalent preference from users.
For practical machine learning systems, performance tuning is often more nuanced than minimizing a single expected loss objective, and it may be more realistically discussed as a multi-objective optimization problem. We propose a novel generative model for scalar-valued utility functions to capture human preferences in a multi-objective optimization setting. We also outline an interactive active learning system that sequentially refines the understanding of stakeholders ideal utility functions using binary preference queries.
This guidebook was written to serve as a useful resource for machine learning practitioners looking to take advantage of Bayesian optimization techniques. We outline four example machine learning problems that can be solved using open source machine learning libraries, and highlight the benefits of using Bayesian optimization in the context of these common applications.
Empirical analysis serves as an important complement to theoretical analysis for studying practical Bayesian optimization. We define two metrics for comparing the performance of Bayesian optimization methods and propose a ranking mechanism for summarizing performance within various genres or strata of test functions. These test functions serve to mimic the complexity of hyperparameter optimization problems, the most prominent application of Bayesian optimization, but with a closed form which allows for rapid evaluation and more predictable behavior.
Bayesian optimization is an elegant solution to the hyperparameter optimization problem in machine learning. Building a reliable and robust Bayesian optimization service requires careful testing methodology and sound statistical analysis. We present an overview of our evaluation system and discuss how this framework empowers our research engineers to confidently and quickly make changes to our core optimization engine
Daily patterns of behaviour are a rich source of information and play an important role in establishing a person’s quality of life. MobiSense is a mobile health research platform that aims to improve mobility analysis for both ambulating and wheelchair users. The goals of the system were to be simple for users to collect mobility data, provide accessible summaries of daily behaviours and to enable further research and development in this area. The system is capable of lifespace summaries relating to indoor and outdoor mobility as well as activity trends and behaviours.
Sublexis is a free tool for improving French, English and Spanish vocabulary. It presents users with chains of flashcards (taken from film scenes) as examples of word context. It's not perfect, but I find it pretty fun for learning new French words myself. I've tried to design it to work well on mobile and touch screen devices, try it out and let me know what you think!
vizdat is an interactive visualization tool that helps users explore and understand their datasets.