Hi, I'm Beck LaBash.

I work with computational systems and like to go fast outdoors.

I'm currently an undergrad at Northeastern, where I compete on the Club Cycling and Nordic Skiing teams. My research focuses on inverse design problems, ocean acidification modeling, and automating software development.

Beck LaBash
$ research
Ocean Acidification
Data-Driven Modeling of 4D Ocean and Coastal Acidification from Surface Measurements - Currently under review
oceanography physics-informed neural nets
RES-Q
A codebase editing benchmark consisting of 100 hand-crafted, compact natural language edit instructions.
code generation benchmarking
$ blog
2024-04-30

Training Generative Flow Networks to select optimal neural network hyperparameters

Developed Generative Flow Networks to sample diverse, high-performing hyperparameter configurations for machine learning models.

reinforcement learning hyperparameter tuning generative flow networks
2024-04-14

lslpy: A python library for contracts, property-based testing, and symbolic execution

lslpy is a library that enables contracts, property-based randomized testing, and verification via symbolic execution in Python, inspired by LSL.

python program verification
2023-12-08

Detecting running form deficiencies with few-shot learning

Fine-tuned a Dual-stream Spatio-temporal Transformer to learn dense representations of running form from monocular video. Can identify potential injury-causing form issues using simple vector classification.

transformers computer vision biomechanics
$ sport
$ benchmarks
track
4:33
1 mile
cycling
326W
FTP
running
16:47
turkey trot
running
1:58
Terminal B->E (Logan Airport)
$ results
3rd/34
L'Enfer du Nord Individual Time Trial
Men's C 2022 — Road Cycling
5th/25
Wild Western Mass Shred-Fest XC
Men's C 2024 — Mountain Biking
$ systm1

YouTube channel where I record me or my friends doing cool stuff outside.