Alan J. Jian

Hi there! I’m a recent graduate of the Data Science program at UC Berkeley. My primary interests lay in applied ML/NLP and Bayesian Statistics.

In just the past year, I’ve had the chance to work on a number of exciting projects, including my capstone project on real-time electricity demand forecasting on CAISO’s energy grid. More recently, I’ve been adapting recent advances in Bayesian Transfer Learning to NLP tasks such as text classification to see if state-of-the-art models could stand to gain from such approaches.

In the past, I’ve also helped deploy a fine-tuned RetinaNet model to automate the counting of elephant seals at Point Reyes, and attempted a slightly more whimsical project that uses BART to create two-sentence horror stories based on posts scrapped from r/TwoSentenceHorror.

I’ve also had the opportunity to TA for the very classes that inspired me to become a data scientist at UC Berkeley, including Data 100, 102, and Data 104. You can read more about those on my teaching page. You can also catch me on Medium, where I occassionally write blog posts on medium about things that I’m working on, or stuff that I’m thinking about.

During my undergraduate (which I also spent at Berkeley), I had the chance to work with Prof. Sara Beckman and Prof. Kosa Goucher-Lambert to study goal formation on student teams using NLP techniques. You can read all about our findings in our conference paper.

Outside of work, you’ll either find me out on the courts playing tennis with friends (always looking for more partners!) or at home playing video games.

If you’re interested in connecting, feel free to reach out to me via LinkedIn, or shoot me an email!

selected publications

IDETC-CIE

Examining Goal Congruence on Engineering Design and Innovation Student Teams

Sara Beckman, Alan Jian, Ahan Sabharwal, and 1 more author

Aug 2021

Abs HTML PDF

Goal congruence, defined as agreement by all members of a team on a common set of objectives, has been positively associated with team cohesion, team performance and team outcomes, including grades earned. Yet there is little in-depth study at scale and across types of engineering design and innovation classes in higher education that examines the goals students set for their work together. This research explores goal congruence in 857 teams involving 1470 students across 18 classes over four years. To examine goal congruence, we use student assessments of their level of agreement on their goals as well as evaluations of their written goal statements. Machine learning techniques are used to automatically identify goal types and congruence between goals. We find that goal congruence on student teams is relatively low, even when they assess it as high, partly due to variety in the types of goals they identify. We categorize the goals students articulate for their teams into grade-, completion-, teaming-, learning-, problem-, output- and outcome-oriented goals and report variance in the types of goals identified in different pedagogical settings. Our findings have implications for how faculty design their classes, link learning outcomes to team projects and facilitate goal setting on student teams.